Petrenko_hw4.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Алгоритмы интеллектуальной обработки больших объемов данных\n",
    "## Домашнее задание №4 - Метод k-средних, предобработка признаков\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### <hr\\>\n",
    "**Общая информация**\n",
    "\n",
    "**Срок сдачи:** 25 апреля 2018, 06:00 <br\\>\n",
    "**Штраф за опоздание:** -2 балла после 06:00 25 апреля, -4 баллов после 06:00 2 мая, -6 баллов после 06:00 9 мая\n",
    "\n",
    "При отправлении ДЗ указывайте фамилию в названии файла<br\\>\n",
    "\n",
    "Используйте данный Ipython Notebook при оформлении домашнего задания.\n",
    "\n",
    "**Штрафные баллы:**\n",
    "\n",
    "1. Невыполнение PEP8 -1 балл\n",
    "2. Отсутствие фамилии в имени скрипта (скрипт должен называться по аналогии со stroykova_hw4.ipynb) -1 балл\n",
    "3. Все строчки должны быть выполнены. Нужно, чтобы output команды можно было увидеть уже в git'е. В противном случае -1 балл\n",
    "4. При оформлении ДЗ нужно пользоваться данным файлом в качестве шаблона. Не нужно удалять и видоизменять написанный код и текст. В противном случае -1 балл\n",
    "<hr\\>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from scipy.spatial.distance import cdist\n",
    "from sklearn.cluster import KMeans\n",
    "import time\n",
    "%matplotlib inline\n",
    "\n",
    "plt.style.use('ggplot')\n",
    "plt.rcParams['figure.figsize'] = (12,5)\n",
    "\n",
    "# %load_ext pycodestyle_magic"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Имплементация K-means\n",
    "\n",
    "Пользуясь наработками выше, имплементируйте метод k-means.  \n",
    "При инициализации необходимо задавать количество кластеров, функцию расстояния между кластерами (для оригинального k-means - евклидово расстояние) и начальное состояние генератора случайных чисел.\n",
    "\n",
    "После обучения, среди атрибутов класса `Kmeans` должны появится\n",
    "* Метки кластеров для объектов\n",
    "* Координаты центройдов кластеров\n",
    "\n",
    "k-means - это алгоритм **кластеризации**, а не классификации, а посему метод `.predict()` в нем фактически не нужен, но он может возвращать метки ближайшего кластера для объектов."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Задание 1 (2 баллов)\n",
    "Имплементируйте метод k-means. Задание считается выполненным, если Ваша реализация работает быстрее реализации из sklearn.\n",
    "\n",
    "Теория для выполнения задания 2 и 3 остаётся на самостоятельное изучение. Теории немного и она совсем простая.\n",
    "\n",
    "###### Задание 2 (2 балла)\n",
    "Имплементируйте класс MiniBatchKMeans, который является классом наследником Kmeans.\n",
    "\n",
    "###### Задание 3 (2 балла)\n",
    "Превратите k-means в k-means++. Для этого нужно реализовать метод в классе Kmeans, который будет инициализировать более \"хорошие\" значения центроидов. Чтобы использовался метод k-means++, в параметр init необходимо передать строковое значение 'k-means' (по умолчанию 'random').\n",
    "\n",
    "###### Задание 4 (2 балла)\n",
    "В пункте \"Проверка корректности метода\" нужно нарисовать графики, которые показывают зависимость времени выполнения алгоритма от количества сэмплов. Графики должны быть  нарисованы для различных комбинаций реализаций алгоритма (k-means, k-means++, k-means с MiniBatchKMeans, k-means++ с MiniBatchKMeans). График достаточно построить на 5-10 точках.\n",
    "\n",
    "###### Задание 5 (2 балла)\n",
    "В пункте \"Применение K-means на реальных данных\" нужно сравнить различные реализации k-means (k-means, k-means++, k-means с MiniBatchKMeans, k-means++ с MiniBatchKMeans). После чего написать вывод, в котором должно быть объяснение того, почему один алгоритм оказался лучше остальных или почему не было выявлено лучшего алгоритма."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "# %%pycodestyle\n",
    "\n",
    "\n",
    "class MyKmeans(object):\n",
    "\n",
    "    def __init__(self, k=2, metric='euclidean', max_iter=1000,\n",
    "                 random_state=None, init='random', epsilon=1e-9):\n",
    "        \"\"\"\n",
    "        Инициализация метода\n",
    "        :k - количество кластеров\n",
    "        :metric - функция расстояния между объектами\n",
    "        :max_iter - максиальное количество итераций\n",
    "        :random_state - seed для инициализации генератора случайных чисел\n",
    "        \"\"\"\n",
    "\n",
    "        self.k = k\n",
    "        self.random_state = random_state\n",
    "        self.metric = metric\n",
    "        self.max_iter = max_iter\n",
    "\n",
    "        self.epsilon = epsilon\n",
    "\n",
    "        if init == 'random':\n",
    "            self.kmeans = False\n",
    "        elif init == 'k-means':\n",
    "            self.kmeans = True\n",
    "\n",
    "    def __init_centroids(self, X):\n",
    "        xmin = np.min(X, axis=0)\n",
    "        xmax = np.max(X, axis=0)\n",
    "\n",
    "        if self.kmeans:\n",
    "            self.centroids[0] = np.random.random_sample((1, X.shape[1]))\\\n",
    "                * (xmax - xmin) + xmin\n",
    "            for j in range(1, self.k):\n",
    "                dists = np.min(cdist(X,\n",
    "                                     self.centroids[:j],\n",
    "                                     metric=self.metric),\n",
    "                               axis=1)\n",
    "                dists = dists / dists.sum()\n",
    "                self.centroids[j] = X[np.random.choice(X.shape[0], p=dists)]\n",
    "        else:\n",
    "            self.centroids = np.random.random_sample((self.k, X.shape[1]))\\\n",
    "                * (xmax - xmin) + xmin\n",
    "\n",
    "    def fit(self, X, y=None):\n",
    "        \"\"\"\n",
    "        Процедура обучения k-means\n",
    "        \"\"\"\n",
    "\n",
    "        # Инициализация генератора случайных чисел\n",
    "        np.random.seed(self.random_state)\n",
    "\n",
    "        # Массив с метками кластеров для каждого объекта из X\n",
    "        self.labels = np.empty(X.shape[0])\n",
    "\n",
    "        # Массив с центройдами кластеров\n",
    "        self.centroids = np.empty((self.k, X.shape[1]))\n",
    "\n",
    "        # Your Code Here\n",
    "        # ...\n",
    "        self.__init_centroids(X)\n",
    "\n",
    "        for i in range(self.max_iter):\n",
    "\n",
    "            # updating labels on current iteration\n",
    "            self.labels = np.argmin(cdist(X,\n",
    "                                          self.centroids,\n",
    "                                          metric=self.metric),\n",
    "                                    axis=1)\n",
    "\n",
    "            # updating centroids\n",
    "            new_centroids = np.empty((self.centroids.shape))\n",
    "            for j in range(self.k):\n",
    "                cluster = X[self.labels == j]\n",
    "                if(cluster.shape[0] != 0):\n",
    "                    new_centroids[j] = cluster.sum(axis=0) / cluster.shape[0]\n",
    "                else:\n",
    "                    new_centroids[j] = self.centroids[j]\n",
    "\n",
    "            # checking whether we can stop\n",
    "            dist = ((self.centroids - new_centroids) ** 2).sum(axis=1)\n",
    "            dist = np.sqrt(dist)\n",
    "            if np.all(dist < self.epsilon):\n",
    "                break\n",
    "\n",
    "            self.centroids = new_centroids\n",
    "        return self\n",
    "\n",
    "    def predict(self, X, y=None):\n",
    "        \"\"\"\n",
    "        Процедура предсказания кластера\n",
    "\n",
    "        Возвращает метку ближайшего кластера для каждого объекта\n",
    "        \"\"\"\n",
    "\n",
    "        return np.argmin(cdist(X, self.centroids, metric=self.metric), axis=1)\n",
    "\n",
    "\n",
    "class MiniBatchKMeans(MyKmeans):\n",
    "    def __init__(self, k=2, metric='euclidean', max_iter=1000,\n",
    "                 batch_size=10, random_state=None,\n",
    "                 init='random', epsilon=1e-3):\n",
    "        super(MiniBatchKMeans, self).__init__(k,\n",
    "                                              metric,\n",
    "                                              max_iter,\n",
    "                                              random_state,\n",
    "                                              init,\n",
    "                                              epsilon)\n",
    "        self.batch_size = batch_size\n",
    "\n",
    "    def __init_centroids(self, X):\n",
    "        xmin = np.min(X, axis=0)\n",
    "        xmax = np.max(X, axis=0)\n",
    "\n",
    "        if self.kmeans:\n",
    "            self.centroids[0] = np.random.random_sample((1, X.shape[1]))\\\n",
    "                * (xmax - xmin) + xmin\n",
    "            for j in range(1, self.k):\n",
    "                dists = np.min(cdist(X,\n",
    "                                     self.centroids[:j],\n",
    "                                     metric=self.metric),\n",
    "                               axis=1)\n",
    "                dists = dists / dists.sum()\n",
    "                self.centroids[j] = X[np.random.choice(X.shape[0], p=dists)]\n",
    "        else:\n",
    "            self.centroids = np.random.random_sample((self.k, X.shape[1]))\\\n",
    "                * (xmax - xmin) + xmin\n",
    "\n",
    "    def fit(self, X, y=None):\n",
    "        # Инициализация генератора случайных чисел\n",
    "        np.random.seed(self.random_state)\n",
    "\n",
    "        # Массив с метками кластеров для каждого объекта из X\n",
    "        self.labels = np.empty(X.shape[0])\n",
    "\n",
    "        # Массив с центройдами кластеров\n",
    "        self.centroids = np.empty((self.k, X.shape[1]))\n",
    "\n",
    "        # Your Code Here\n",
    "        # ...\n",
    "        self.__init_centroids(X)\n",
    "\n",
    "        for i in range(self.max_iter):\n",
    "\n",
    "            # updating labels on current iteration\n",
    "            self.labels = np.argmin(cdist(X,\n",
    "                                          self.centroids,\n",
    "                                          metric=self.metric),\n",
    "                                    axis=1)\n",
    "            mbatch_idx = np.random.choice(X.shape[0],\n",
    "                                          size=self.batch_size,\n",
    "                                          replace=False)\n",
    "            mbatch = X[mbatch_idx]\n",
    "            mlabels = self.labels[mbatch_idx]\n",
    "            # updating centroids\n",
    "            new_centroids = np.empty((self.centroids.shape))\n",
    "            for j in range(self.k):\n",
    "\n",
    "                cluster = mbatch[mlabels == j]\n",
    "\n",
    "                if(cluster.shape[0] != 0):\n",
    "                    new_centroids[j] = cluster.sum(axis=0) / cluster.shape[0]\n",
    "                else:\n",
    "                    new_centroids[j] = self.centroids[j]\n",
    "\n",
    "            # checking whether we can stop\n",
    "            dist = ((self.centroids - new_centroids) ** 2).sum(axis=1)\n",
    "            dist = np.sqrt(dist)\n",
    "            if np.all(dist < self.epsilon):\n",
    "                break\n",
    "            self.centroids = new_centroids\n",
    "        return self"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Проверка корректности метода\n",
    "\n",
    "Перед тем как применять алгоритм на реальных данных, нужно испытать его на простых \"игрушечных\" данных.\n",
    "\n",
    "Если алгоритм реализован правильно, то метод должен идеально разбивать на 3 кластера данные ниже. Проверьте это.\n",
    "\n",
    "ВНИМАНИЕ! Проверка должна быть осуществлена на всех реализациях, иначе реализация не будет зачтена!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.datasets import make_blobs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [],
   "source": [
    "X, y = make_blobs(n_samples=100, n_features=2, centers=3, cluster_std=1, \n",
    "           center_box=(-10.0, 10.0), shuffle=False, random_state=1234)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x10b142c50>"
      ]
     },
     "execution_count": 80,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(X[:,0], X[:, 1], c=y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Проверьте Вашу имплементацию на простых данных (без этого пункта ДЗ не считается выполненным).   \n",
    "КОММЕНТАРИИ НЕ СТИРАТЬ!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x10dad1e50>"
      ]
     },
     "execution_count": 82,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Работоспособность MyKmeans\n",
    "km = MyKmeans(k=3, init='random')\n",
    "km.fit(X)\n",
    "plt.scatter(X[:, 0], X[:, 1], c=km.labels)\n",
    "plt.scatter(km.centroids[:, 0], km.centroids[:, 1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x10db5dad0>"
      ]
     },
     "execution_count": 83,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Работоспособность MyKmeans++\n",
    "kmpp = MyKmeans(k=3, init='k-means')\n",
    "kmpp.fit(X)\n",
    "plt.scatter(X[:, 0], X[:, 1], c=kmpp.labels)\n",
    "plt.scatter(kmpp.centroids[:, 0], kmpp.centroids[:, 1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mbkm quitting on iteration 999\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x1a21021410>"
      ]
     },
     "execution_count": 85,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Работоспособность MyKmeans с MiniBatchMyKmeans\n",
    "mbkm = MiniBatchKMeans(k=3, init='random', batch_size=10)\n",
    "mbkm.fit(X)\n",
    "plt.scatter(X[:, 0], X[:, 1], c=mbkm.labels)\n",
    "plt.scatter(mbkm.centroids[:, 0], mbkm.centroids[:, 1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mbkm quitting on iteration 999\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x1a21107150>"
      ]
     },
     "execution_count": 86,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Работоспособность MyKmeans++ с MiniBatchMyKmeans\n",
    "mbkmpp = MiniBatchKMeans(k=3, init='k-means', batch_size=10)\n",
    "mbkmpp.fit(X)\n",
    "plt.scatter(X[:, 0], X[:, 1], c=mbkmpp.labels)\n",
    "plt.scatter(mbkmpp.centroids[:, 0], mbkmpp.centroids[:, 1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1a1f4faa50>]"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Время выполнения алгоритма Kmeans из sklearn\n",
    "\n",
    "skm = KMeans()\n",
    "skmtimes = np.zeros(10)\n",
    "iters = 10\n",
    "size_grid = np.linspace(1000, 10000, 10, dtype=int)\n",
    "for i, size in enumerate(size_grid):\n",
    "    X, y = make_blobs(n_samples=size, n_features=2, centers=5, cluster_std=1, \n",
    "           center_box=(-10.0, 10.0), shuffle=False, random_state=1337)\n",
    "    for j in range(iters):\n",
    "        t1 = time.time()\n",
    "        skm.fit(X)\n",
    "        skmtimes[i] += time.time() - t1\n",
    "    skmtimes[i] /= iters\n",
    "plt.plot(size_grid, skmtimes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1a214ea110>]"
      ]
     },
     "execution_count": 91,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Время выполнения алгоритма MyKmeans\n",
    "\n",
    "skm = MyKmeans(k=5, init='random')\n",
    "kmtimes = np.zeros(10)\n",
    "iters = 10\n",
    "size_grid = np.linspace(1000, 10000, 10, dtype=int)\n",
    "for i, size in enumerate(size_grid):\n",
    "    X, y = make_blobs(n_samples=size, n_features=2, centers=5, cluster_std=1, \n",
    "           center_box=(-10.0, 10.0), shuffle=False, random_state=1337)\n",
    "    for j in range(iters):\n",
    "        t1 = time.time()\n",
    "        skm.fit(X)\n",
    "        kmtimes[i] += time.time() - t1\n",
    "    kmtimes[i] /= iters\n",
    "plt.plot(size_grid, kmtimes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1a21698350>]"
      ]
     },
     "execution_count": 93,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYoAAAD8CAYAAABpcuN4AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzt3XmcFMX5+PFPzS63AsJ6cKkoeACKBEXUeIEiRgWN5gE1KkYlmBBvEzWSqNEETSKS6NdfEA9EIz5iVJDLCEjwQhCNBjw4xWVVZBcBOXbZnfr90b0yO+4xsEfP7D7v12teTvdU11SXzT5T1dVVznuPMcYYU5FY1AUwxhiT3ixQGGOMqZQFCmOMMZWyQGGMMaZSFiiMMcZUygKFMcaYSlmgMMYYUykLFMYYYyplgcIYY0ylsqMuQA2xx8uNMWb3uKoS1JdAQV5eXtRFqJacnBzWr18fdTHShtXHTlYXZVl9lFWd+mjfvn1K6azryRhjTKUsUBhjjKmUBQpjjDGVskBhjDGmUhYojDHGVMoChTHGmEpZoDDGGFMpCxTGGJOBfDxO/LnHKP58Va1/lwUKY4zJQP7dN/CvvMiOZR/V+ndZoDDGmAzji3fgX5gIHQ+k6cln1Pr3WaAwxpgM4+fNhK+/JHb+MFxWVq1/nwUKY4zJIH7rFvzLk+DwntC9V518Z0qTAorIQGAskAWMV9XRSZ83AZ4EegP5wBBVXS0ipwOjgcZAEXCzqs4RkT2B+QlZdASeUtXrRGQY8GdgbfjZg6o6fndP0Bhj6hM/83n4dnPQmnBVTvxaI6oMFCKSBTwEnA7kAgtFZIqqLk1IdgWwQVW7iMhQ4F5gCLAeOEdV80SkBzAL6KCqm4GjEr7jXeBfCfk9q6ojq3luxhhTr/iC9fhXp+COPRl3wMF19r2pdD31AZar6kpVLQImAYOT0gwGJoTvJwP9RcSp6nuqWjr/9xKgadj6+I6IdAX2oWwLwxhjTBI/5Wnwcdy5P63T700lUHQAPk/Yzg33lZtGVYuBjUDbpDTnA++pamHS/gsJWhCJiw+dLyIfiMhkEemUQhmNMaZe87mr8W/OwfU7G5ezb51+dyr3KMrrBEteUa7SNCLSnaA7akA56YYClyRsTwWeUdVCERlB0FLpl3yQiAwHhgOoKjk5OZWdQ9rLzs7O+HOoSVYfO1ldlNVQ62PD//sTO5rvQc5PRxDbs+V3++uiPlIJFLlA4q/6jkDycnKlaXJFJBtoBRQAiEhH4AXgUlVdkXiQiPQEslX13dJ9qpqfkOQRggDzPao6DhgXbvpMX/HKVu0qy+pjJ6uLshpiffiP/kv83bdwF1xOQWERFO48/3RZ4W4h0FVEOotIY4IWwJSkNFOAy8L3FwBzVNWLSGtgGnCrqr5RTt4XAs8k7hCRdgmbg4Daf+zQGGPSlI/HiT8/Adrsjet3ViRlqLJFoarFIjKSYMRSFvCYqi4RkbuARao6BXgUmCgiywlaEkPDw0cCXYBRIjIq3DdAVdeF7wX4UdJXXiMig4DiMK9hu312xhiT4fzC+fDZctzPrsc1ahxJGZz3ybcbMpLPy0vuDcssDbE5XRmrj52sLspqSPXhd+wgPupqaNaC2KgxuNj3O4FqoOupyocx7MlsY4xJU/616ZC/jthPhpUbJOqKBQpjjElDfuu3+GkK3XrhutXNVB0VsUBhjDFpyM94HrZ+S+z8y6pOXMssUBhjTJrx+V+HU3Wcgtv/oKiLY4HCGGPSjX/paQDcuRdHXJKABQpjjEkj/vNV+Lfn4vqfjWu7T9TFASxQGGNMWok//wQ0a4E78ydRF+U7FiiMMSZN+KXvw5L3cGcJrsUeURfnOxYojDEmDQRTdTwBbffBnRrNVB0VsUBhjDFpwL/zH1izEnfeJbhGjaIuThkWKIwxJmJ+RxH+xadg/4Nxx5wYdXG+xwKFMcZEzM+dFkzVcUG0U3VUJP1KZIwxDYjfsjmYqqPHD3CH94y6OOWyQGGMMRHy0yfDtq1pMVVHRSxQGGNMRHz+Ovycqbjj+uE6do66OBVKZSlURGQgMJZg4aLxqjo66fMmwJNAbyAfGKKqq0XkdGA00BgoAm5W1TnhMa8B7YBtYTYDVHVdRXlV5ySNMSYd+RefBhfDDb4o6qJUqsoWhYhkAQ8BZwLdgAtFpFtSsiuADaraBRjDznWu1wPnqOoRBEulTkw67mJVPSp8rasiL2OMqTf8mhX4Ba/h+p+Da7N31MWpVCpdT32A5aq6UlWLgEnA4KQ0g4EJ4fvJQH8Rcar6nqqWLj23BGgathgqU25eKZTTGGMyRvz5CdB8D9yZ50ddlCqlEig6AJ8nbOeG+8pNo6rFwEagbVKa84H3VLUwYd/jIvK+iIxKCAap5GWMMRnLL3kPlr6PO1twzdNnqo6KpHKPorxf88kLbVeaRkS6E3QhDUj4/GJVXSsiewLPA5cQ3JtI5fsQkeHAcABVJScnp7JzSHvZ2dkZfw41yepjJ6uLsjK9PnxJCQUvPoXbtz1tz78E16hxtfKri/pIJVDkAp0StjsCeRWkyRWRbKAVUAAgIh2BF4BLVXVF6QGqujb872YR+SdBF9eTleWVSFXHAePCTZ/pi603pAXjU2H1sZPVRVmZXh/xN+fgVy/DXXUT+Rs3VTu/6tRH+/btU0qXSqBYCHQVkc7AWmAokHyLfgrBzeq3gAuAOarqRaQ1MA24VVXfKE0cBoDWqrpeRBoBZwOvVpZXSmdjjDFpzO8owr/0FBzQBXf0D6MuTsqqDBSqWiwiI4FZBMNjH1PVJSJyF7BIVacAjwITRWQ5wa//oeHhI4EuwCgRGRXuGwBsAWaFQSKLIEg8En5eUV7GGJPR/JyXoWA9scuvS8upOirivK8XP9Z9Xl5yb1hmyfTmdE2z+tjJ6qKsTK0P/+0m4rf9HLocTtY1v6uxfGug66nKUaWZE9KMMSaD+enPwfZtaT1VR0UsUBhjTC3zX3+JnzsNd3w/XIcDoi7OLrNAYYwxtcy/+DTEYrhB6T1VR0UsUBhjTC3yny3HvzMPd9pgXJvMfP7DAoUxxtQS7z3xyU/AHi1xZ/w46uLsNgsUxhhTW5Ysho8/wJ09BNe8RdSl2W0pTTNujGm4/Cf/Iz7tWfh8FXQ+BNe1O+6Q7nDAwbjsRlEXL235eEnQmth7P9zJA6MuTrVYoDDGfI/3Hj7+gPjLk+DTJdCyNa7HD/Crl+E/XBRMvta4MXQ+NAgcXbvBQYfimjaLuuhpw7/1Gqz9DDf81xkfUC1QGGO+472HJYuJv/wsrPgYWrfBDb0Kd+IAXONghQC/aQMs+wi/bAl+2VL8NMX7OMRiwdQUXbsFgaNLN9weLSM+o2j4okL8i0/BgV1xR58QdXGqzQKFMSYIEB8sJD51Eny2HNrk4C4egTvhtO/Nbupa7gW9j8f1Pj44dttWWPFREDQ+XYKf8zL+lReDxO33D4JG1+5ByyNDR/3sKj97KnyTT+zKG3Eu85fTsUBhTAPm43F4/23i0xTWrIScfXGX/DJ4MCzF7hLXrDn06I3r0TvIc0cRrFoWtjiW4BfMg3kzg+6qtvvgunaHQ8Luqn071Is/pIn85k34GZOhZx/coT2iLk6NsEBhTAPk4yX4d9/ET1NY+xns0w437FrcsSfjsqv3Z8E1ahwEgkO6B99VUgK5q3Z2VS1ZDG/PDQLHnq3C1ka3IIB0OhAXy6r+CUbIT1fYvp3Yjy+Nuig1xgKFMQ2ILynBL5wfBIgvc2G/jrgrbsAdcyIuq3b+QLusrODexQFd4LTBQTfXl2vxy5bAsqVBAFn8ZhA4mjWHgw/DdemGO6RH0MffKHNuBAdTdUzH/fA0XPv9oy5OjbFAYUwD4IuL8QvmBb92130BHQ4IRuP0Pq7Of8E756BdR1y7jnDSGUH5Cr7GL1sKn4bdVf97Kggc2Y2gc1dc1x7hDfLD6rSsu8q/MBGyYrhBF0ZdlBplgcKYeswX78C/OSfoM1//FXTqTOzqW+Covmm1HoJrszfu2JPh2JOBoJ+f5Ut3dlfNnIyfHgcXo6Dr4cS79cIddWxwszxN7nH4VcvwC+fjzhJc67ZRF6dGWaAwph7yO4rwr7+KnzkZCtbDgV2JDb0Kjjwmbf6wVsbt2RJ69cX16guA374NVn4cjKr65EP8i08Fw09z9sX17IPr2Se411HN+yu7K5iq43HYs1VGT9VRkZRqVUQGAmMJVqMbr6qjkz5vQrDedW8gHxiiqqtF5HRgNNAYKAJuVtU5ItIceA44GCgBpqrqLWFew4A/Eyy7CvCgqo6v1lka00D4okL8/FfwM5+Hbwrg4MOIXfJL6P6DjAgQFXFNm0G3XrhuvWibk8PXyz/Bf7AQ//47+Hkzg+GozVrgjugdjDbq8QNc8z3qroAfLoJP/4e76OfBKLB6pspAISJZwEPA6UAusFBEpqjq0oRkVwAbVLWLiAwF7gWGAOuBc1Q1T0R6ECyn2iE85i+qOldEGgOzReRMVZ0Rfvasqo6skTM0pgHwhdvx82bgZ70Am76BQ7oT+9n1cNiRGR0gKuJat8WdNBBOGogv3A5L38f/dwH+g0Xwzn/wWVlwSI+gtXHkMbi996u1svh4CfHnJwQjx048o9a+J0qptCj6AMtVdSWAiEwCBgOJgWIwcEf4fjLwoIg4VX0vIc0SoKmINFHVrcBcAFUtEpHFQMdqnYkxDZDfvhU/d3rwgNu3m+DwnsSG/7rejN9PhWvS9LtuKh8vgZWf4v/7TvCa9Ah+0iPBzfvSLqoDu9bo/Rn/5hzIW0NsxG8i6/qqbamcVQfg84TtXODYitKoarGIbATaErQoSp0PvKeqhYkHikhr4ByCrq3v0orIScCnwPWqmvj9pccNB4aH30lOTmY/8ZmdnZ3x51CTrD52Kq8u4ls2s3XaZLZOfRb/7SYa9+pLC7mcxocdEVEp606V18Y++0LfEwEo/iKXwoWvU7jwdXbM/Bd++nPEWreh8dEn0KTPiTQ+8ugg0OwmX7id9VOfIfuQ7uw1YFAkrbe6+LeSSqAo78z9rqQRke4E3VEDEhOISDbwDPC30hYLMBV4RlULRWQEMAHol5y5qo4DxpV+VyYutp4oUxeMry1WHzsl1oXfshn/6tSgT37bFujZh9hZQknnQ9gE0ADqbJeujUZN4fjT4PjTiG3ZjP/wXfx/32Hb66+y7dWpwcSGhx8VtjaOCaYn2QXxaYovWA9X3Eh+fv5unE31VeffSvv27VNKl0qgyAU6JWx3BPIqSJMb/vFvBRQAiEhH4AXgUlVdkXTcOGCZqj5QukNVE2v7EYIAY0yD5jdvwv/7RfzcabB9G/TqS+ysIbgDDo66aBnDtdgT1/cU6HsKvngHfPq/4GZ4aTeVc8E06j374HoeC+07VdpC8Js3BoMGjjr2u6fQ66tUAsVCoKuIdCYYiTQUSF74dQpwGfAWcAEwR1V92K00DbhVVd9IPEBE7iYIKFcm7W+nql+Em4OAj3btlIypP/y2rWye8BDxGc9DUSGu9wnBOP2OB0ZdtIzmsht9N4rKXzgc1q7eGTRemBg8OFfF0Fv/8rNQVEjsx5dFdBZ1p8pAEd5zGEkwYikLeExVl4jIXcAiVZ0CPApMFJHlBC2JoeHhI4EuwCgRGRXuG0AwXPa3wMfAYhGBncNgrxGRQUBxmNewGjlTYzKM37KZ+AN3sHXNimCKjbME165T1QeaXeKcg46dcR07w9lD8N/kVzn0lm834efNwP1wQPCEeT3nvE++3ZCRfF5ecm9YZrE++bIaen34Td8QH/M7+DKXVjffw7cHHR51kdJGXV4b3xt6u3kjZGUFkxlu3ULsnn/gWrepk7JUpAbuUVR5B75+juUyJoP5DfnE7x8FBeuIjRxF0z4n8m0DDppR+t7Q21XLgqDxv8W4M34ceZCoKxYojEkjPn8d8b/eDps2Erv2jmAGVZMWXCwrmNn24MOgAdyXSGSBwpg04b/KI37/7bB9G7Eb7sIddGjURTIGsEBhTFrwa9cQHzMKSkqI3XgPbv+Doi6SMd9Jn3mGjWmg/JoVxP9yGwCxm/9oQcKkHWtRGBMhv/IT4mPvgKbNiN1wN27f1J6UNaYuWaAwJiL+k/8R//sfoGUrYjfejWu7T9RFMqZcFiiMiYBf8h7x/7sH2uxD7MY/1LsV0Uz9YoHCmDrm319A/B/3wn6diF1/J65l66iLZEylLFAYU4fiC+fjH70fOh1E7Lo7cC32jLpIxlTJAoUxdST+5mz8E38Plie95nf1cslMUz9ZoDCmDsRfm4F/+uFgBbpf/rZai+UYU9csUBhTy+KvvIh/7jE48phgucxGjaMukjG7xAKFMbXEe4+fpviXng7WkbjyhmAdBGMyjAUKY2qB9z5YAGfGZFzfU3HDrsFlZUVdLGN2S0qBQkQGAmMJFi4ar6qjkz5vAjwJ9AbygSGqulpETgdGEyxUVATcrKpzwmN6A08AzYDpwLXhqnhtgGeBA4HVgKjqhuqdpjF1x3uPf3Y8fvZU3Eln4C6+Ghez2XJM5qry6hWRLOAh4EygG3ChiHRLSnYFsEFVuwBj2LnO9XrgHFU9gmCp1IkJxzwMDAe6hq+B4f5bgNmq2hWYHW4bkxF8vAQ/8aEgSJw2CPfTX1iQMBkvlRZFH2C5qq4EEJFJwGBgaUKawcAd4fvJwIMi4lT1vYQ0S4CmYeujDdBSVd8K83wSOBeYEeZ1SnjMBOA14De7emLG1DVfUoJ/Yiz+7ddwPxLcuRcHy2wak+FS+anTAfg8YTs33FduGlUtBjYCyXMSnA+8p6qFYfrcCvLcV1W/CPP6ArAJcEza88U7iI/7cxAkzv0psfN+akHC1BuptCjKu9qTF9quNI2IdCfojhqwC3lWSkSGE3Rdoark5OTsyuFpJzs7O+PPoSZlUn34okK+ue+3FC1+kz1+di0tzhlSo/lnUl3UBauPsuqiPlIJFLlAp4TtjkBeBWlyRSQbaAUUAIhIR+AF4FJVXZGQvmMFeX4lIu1U9QsRaQesK69QqjoOGBdu+rpabL221OWC8ZkgU+rDF24n/tA98PEHuJ/+gm3H9WdbDZc7U+qirlh9lFWd+mjfPrVp7VPpeloIdBWRziLSGBgKTElKM4XgZjXABcCccARTa2AacKuqvlGaOOxS2iwifUXEAZcCL5WT12UJ+41JK37bVuIP3AEff4gbdi2xkwdWeYwxmajKQBHecxgJzAI+CnbpEhG5S0QGhckeBdqKyHLgBnaOVBoJdAFGicj74av0nsPVwHhgObCC4EY2BMNpTxeRZUDp8Fpj0orfspn4X2+HVZ8QG34TseP7RV0kY2qN836Xbg2kK5+Xl9wbllmsOV1WOteH3/QN8TG/gy9ziY24BdezT61+XzrXRRSsPsqqga6nKkdd2JPZxuwCvyGf+P2joGAdsV+NwnXrFXWRjKl1FiiMSZFf/1UQJDZtJHbtnbhDukddJGPqhAUKY1Lgv8ojfv/tsH1bsHRp50OiLpIxdcYChTFV8GvXEB8zCkpKiN14D27/g6IukjF1ygKFMZXwa1YEN66zGhH79Z9w7TpVfZAx9YwFCmMq4Fd8THzsndCsedDdtE9qDycZU99YoDCmHP6zFcTH/B5atiJ24924tjblmGm4LFAYk8R/U0D8wbuhRYugu6l18vyWxjQsNlG+MQl8UWEwd9O2LcRGjrIgYQwWKIz5jvce/8Tf4LPlxK68Adepc9RFMiYtWKAwJuSnPYtfOB933qW4o/pGXRxj0oYFCmMAv+h1/Ev/xB13Km7gj6MujjFpxQKFafD8Z8uJP/4AHHwY7pKRtjKdMUksUJgGzX+TH4xw2qMVsV/chmvUKOoiGZN2LFCYBssXFhJ/8B7Yto3Yr27HtWwddZGMSUspPUchIgOBsUAWMF5VRyd93gR4EugN5ANDVHW1iLQFJgPHAE+o6sgw/Z7A/IQsOgJPqep1IjIM+DOwNvzsQVUdv5vnZ0y5ghFOY2HNCmK//C2uo41wMqYiVQYKEckCHiJYbS4XWCgiU1R1aUKyK4ANqtpFRIYC9wJDgO3AKKBH+AJAVTcDRyV8x7vAvxLye7Y0qBhTG/zUSfhFr+MuGFbrCw8Zk+lS6XrqAyxX1ZWqWgRMAgYnpRkMTAjfTwb6i4hT1S2q+jpBwCiXiHQF9qFsC8OYWhNfOB8/9Rnccf1wA86LujjGpL1UAkUH4POE7dxwX7lpwjW2NwKpPtJ6IUELInFN1vNF5AMRmSwiNl2nqTF+1TL842Ohy+G4S35pI5yMSUEq9yjK+5eUvNB2KmkqMhS4JGF7KvCMqhaKyAiClsr3Vq4XkeHAcABVJScnJ8WvS0/Z2dkZfw41qTbqoyT/awoe/hOxvdrS9va/EGu1V43mX1vs2ijL6qOsuqiPVAJFLpD4q74jkFdBmlwRyQZaAQVVZSwiPYFsVX23dJ+q5ickeYTgfsf3qOo4YFy46TN9sXVbML6smq4PX1hI/L5bYOsWYrfeR8GOEsiQ+rZroyyrj7KqUx/t26c2dX4qXU8Lga4i0llEGhO0AKYkpZkCXBa+vwCYk9SVVJELgWcSd4hIu4TNQcBHKeRjTIV8PE788THw+UpiV92E63BA1EUyJqNU2aJQ1WIRGQnMIhge+5iqLhGRu4BFqjoFeBSYKCLLCVoSQ0uPF5HVQEugsYicCwxIGDElwI+SvvIaERkEFId5DavG+RmDnzoJ3n0T95PLcT2Pibo4xmQc532qtxLSms/LS+4NyyzWnC6rpuojvmAefvxfcSf0x112TUbevLZroyyrj7JqoOupyn8U9mS2qbf8qk+DacO7dsNd/IuMDBLGpAMLFKZe8gXrgwWIWrchdvWtNoeTMdVggcLUO75wO/GH7obC7cEqdXu2irpIxmQ0CxSmXvHxOPHHxsDnq4kNvxnXYf+oi2RMxrNAYeoVP+WfsPitYA6nI46OujjG1AsWKEy9EV8wDz9NcT88HXd68nRkxpjdZYHC1At+5SfBCKdDuuMuHmEjnIypQRYoTMbzBV8HI5z2aktsxK24bBvhZExNskBhMprfvo343++GHUXERt6O27Nl1EUypt6xQGEylo/HiT86BtZ+Ruyqm3HtbYSTMbXBAoXJWP6lp+H9t3FyOe6I3lEXx5h6ywKFyUjxt+fipz+HO3EArv+gqItjTL1mgcJkHL/iY/yEB+HQI3AX/dxGOBlTyyxQmIzi89cljHD6jY1wMqYOpLLCnTFpwW/fRvzBu6F4B7Ff/RG3h41wMqYuWIvCZIRghNP9sHZNMIdTu05VH2SMqREptShEZCAwlmCFu/GqOjrp8ybAk0BvIB8YoqqrRaQtMBk4BnhCVUcmHPMa0A7YFu4aoKrrKsprt8/Q1Av+xYnw/gLc0KtwPWyEkzF1qcoWhYhkAQ8BZwLdgAtFpFtSsiuADaraBRgD3Bvu3w6MAm6qIPuLVfWo8LWuirxMAxV/cw5+xvO4kwbi+p0ddXGMaXBS6XrqAyxX1ZWqWgRMApJnXBsMTAjfTwb6i4hT1S2q+jpBwEhVuXntwvGmHvHLP8JPDEc4XTjcRjgZE4FUup46AJ8nbOcCx1aURlWLRWQj0BaoaiHXx0WkBHgeuFtVfap5ichwYHiYjpycnBROJX1lZ2dn/DnUpOzsbPaK7yD/4T+Rtfd+tLntPmItG+YCRHZtlGX1UVZd1EcqgaK8n3B+N9Iku1hV14rIngSB4hKCexMp5aWq44BxpZ9n+mLrtmB8WW1aNGP9XTfAjh34q2+joGgHNND6sWujLKuPsqpTH+3bt08pXSpdT7lA4hCTjkBeRWlEJBtoBRRUlqmqrg3/uxn4J0EX127lZeoXHy9h05g74YvPif3817h2HaMukjENWiotioVAVxHpDKwFhgIXJaWZAlwGvAVcAMwJu5HKFQaA1qq6XkQaAWcDr+5OXqb+8f+aSOHC14N7Et17RV0cYxq8KgNFeJ9gJDCLYHjsY6q6RETuAhap6hTgUWCiiCwn+PU/tPR4EVkNtAQai8i5wADgM2BWGCSyCILEI+EhFeZl6r/4qy/hZ/2LZmecR+GpZ0VdHGMM4LyvFz/WfV5ecm9YZrF+V4i/8WqwSt0PjmOf2+4lf8M3URcpLdi1UZbVR1k1cI+iyqGE9mS2SQt+8ZvBRH/djiJ25U24LJtdxph0YYHCRM4vfY/4I3+Bgw4h9ovbcI1soj9j0okFChMpv+Jj4g/9EfbrSOxXv8M1aRp1kYwxSSxQmMj43FXE/3YntG5D7Lo7cS32iLpIxphyWKAwkfBf5REf83to3JTY9XfhWu0VdZGMMRWwO4amzvmC9cTH/A7icWI33YPL2TfqIhljKmGBwtQpv3ljECS2bA6ChK0rYUzas64nU2f8tq3Ex94J+euIjRyFO6BL1EUyxqTAAoWpE76okPiDf4DcVcFa14f2iLpIxpgUWaAwtc4XFxP/f/fCsqW4y6/DHXlM1EUyxuwCCxSmVvl4Cf7xB+DDRbiLRhA79uSoi2SM2UUWKEyt8d7j//kP/Dv/wf34UmKnnBl1kYwxu8EChak1/oWJ+HkzcWf8mNiZF0RdHGPMbrJAYWpFfObz+BmTcSedgTv/sqiLY4ypBgsUpsbF/zMT//wE3DEn4i4egXNVzmJsjEljKT1wJyIDgbEEiwyNV9XRSZ83IVjvujeQDwxR1dUi0haYDBwDPKGqI8P0zYHngIOBEmCqqt4SfjYM+DPBanoAD6rq+OqcpKk78YXz8U89DD164352HS6WFXWRjDHVVGWLQkSygIeAM4FuwIUi0i0p2RXABlXtAowB7g33bwdGATeVk/VfVPUwoBdwgogk3ul8VlWPCl8WJDKE/3AR/tH7ocvhxEbcgsu26cKNqQ9S6XrqAyxX1ZWqWgRMAgYnpRkMTAjfTwb6i4hT1S2q+jpBwPiOqm5V1bnh+yJgMdCxGudhIuY/XUL84dHQ4YDgqesmTaIukjGmhqQSKDoAnyds54b7yk2jqsXARqBtKgUQkdbAOcDshN3ni8gHIjJZRGwyoDTnP1sRPHXddu9guvDmLaIukjGmBqVyj6K8O5HJC22nkuZ7RCQbeAb4m6quDHdPBZ5R1UIRGUFSWqcCAAAOsUlEQVTQUulXzrHDgeEAqkpOTk5VX5fWsrOzM/Icitd+RsHf7iS2x560+cODZNXQTLCZWh+1weqiLKuPsuqiPlIJFLlA4q/6jkBeBWlywz/+rYCCFPIeByxT1QdKd6hqfsLnj7DzfkcZqjouPB7AZ/pi65m4YLzP/5r4fb8B73HX3MEGsqCGziET66O2WF2UZfVRVnXqo3379imlS6XraSHQVUQ6i0hjYCgwJSnNFKB0sPwFwBxVrbRFISJ3EwSU65L2t0vYHAR8lEIZTR3zmzYQv38UbNsWLDy0X3JvpDGmvqiyRaGqxSIyEphFMDz2MVVdIiJ3AYtUdQrwKDBRRJYTtCSGlh4vIquBlkBjETkXGABsAn4LfAwsFhHYOQz2GhEZBBSHeQ2roXM1NcRv/Zb4mDvgm/VBkOjUOeoiGWNqkfO+ylsJmcDn5SX3hmWWTGlO+8LtxB/4PaxaRmzk7bgeP6iV78mU+qgLVhdlWX2UVQNdT1U+EWtPZpuU+eIdxB/+E6z4hNhVN9ZakDDGpBcLFCYlPl6CH38/LHkPd8kvcL1PiLpIxpg6YoHCVMl7j5/4f/h338D95HJiJw6IukjGmDpkgcJUynuPn/w4/vV/484SYgPOi7pIxpg6ZoHCVMpPfw7/you4U8/CDb446uIYYyJggcJUKD53Gv7Fp3B9T8ENvcqmCzemgbJAYcoVf3su/p//gJ59cJddg4vZpWJMQ2X/+s33+PcX4B8fC4ceQeznv8Zlp7RsiTGmnrJAYcrwH39A/B/3wf4HExv5W1yjxlEXyRgTMQsU5jt+5SfEH7wH9t6P2LW/xzVtHnWRjDFpwPoUGjjvPXz8AfHZU+GDhdB2H2I33IXbo2XURTPGpAkLFA2ULyzEL5iLn/0y5K2BPVrizvwJrv/ZuJatoy6eMSaNWKBoYHz+Ovzcafj5/4at30Knzrhh1+L6nGj3I4wx5bJA0QB47+HTJcTnTIX3FgRzRfbqS6z/IOhyuD0fYYyplAWKeswXFeLf+U/QvZS7ClrsiRt4Hu7kH+Ha7h118YwxGSKlQCEiA4GxBAsXjVfV0UmfNwGeBHoD+cAQVV0tIm2BycAxwBOqOjLhmN7AE0AzYDpwrap6EWkDPAscCKwGRFU3VOMcGxxfsB7/2nT8/Fnw7WbocADu0pG4Y0/GNW4SdfGMMRmmyuGxIpIFPAScCXQDLhSRbknJrgA2qGoXYAw717neDowCbion64eB4UDX8DUw3H8LMFtVuwKzw21TBe89fvlS4v+4j/itV+Jn/gu6dCd20z3Efv83YicOsCBhjNktqbQo+gDLVXUlgIhMAgYDSxPSDAbuCN9PBh4UEaeqW4DXRaRLYobhutgtVfWtcPtJ4FxgRpjXKWHSCcBrwG929cQaCr9jB35h2L20ZgU0b4E7bTDulDNxe+8XdfGMMfVAKoGiA/B5wnYucGxFacI1tjcCbYGK1ufrEOaTmGeH8P2+qvpFmNcXIrJPCmVscPw3+fh5M/HzZsLmjdCuE+7iq3HHnYpr0jTq4hlj6pFUAkV5Q2KSF9pOJU110n+PiAwn6LpCVcnJydmVw9NOdnZ2Suew49MlbH1Z2f7mHIjHaXz0CTQ/6yc0PvLoejV6KdX6aAisLsqy+iirLuojlUCRC3RK2O4I5FWQJldEsoFWQEEVeXasIM+vRKRd2JpoB6wrLwNVHQeMCzd9pi+2XtkC6b54B37RG/g5L8OqT6FZ82B9iFPPomSfdmwGyM+v0/LWtuosGF/fWF2UZfVRVnXqo3379imlSyVQLAS6ikhnYC0wFLgoKc0U4DLgLeACYI6qVthCCIPAZhHpCywALgX+npTX6PC/L6V0JvWQ37QBP28Wft4M2LgB9u2Au+jnQfeSzcNkjKkjVQaK8J7DSGAWwfDYx1R1iYjcBSxS1SnAo8BEEVlO0JIYWnq8iKwGWgKNReRcYICqLgWuZufw2BnhC4IAoSJyBbAG+ElNnGgm8auX4We/jF80H4qLoUdvYsPOhm69bF0IY0ydc97v0q2BdOXz8pJ7wzJL29at+frfU/Gzp8KKj6FJM9wJ/YMupv06VJ1BPWPdCztZXZRl9VFWDXQ9VXlz057Mjpj/pgD/+r9Z//or+PyvYe/9cEOuxB3fH9e8RdTFM8YYCxRR+G5q73kz4P0FUFJCo57HUHzhCDiit3UvGWPSigWKOuS3fIt/c3bw7MNXa4O5l/qfgztpIHt1P9Ka08aYtGSBopZ572H1MvxrM/AL58OOIjj4MNxZ1+N6H2/Tahhj0p4FilriC7fjF8wLhrauWQlNmuKO6xdMrdGpc9TFM8aYlFmgqGF+7Rr8vBn4t+fCtq3BzK0Xj8AdewqumT37YIzJPBYoaoDfsQO/+M2g9bBsKWRn447+Ie7kgXCwLQxkjMlsFiiqwX/9JX7+LPzrrwYT8+29H+6CYbjjT8Pt2TLq4hljTI2wQLGLfLwEPnyX+GszYMliwEHPPsROHgjdjrKhrcaYescCRYr8xg34+a8Eq8YVrIdWbXBnDcGdOADXxmayNMbUXxYoKuG9h08+DIa2vv82lJTA4T2JDbkSjuyDy7bqM8bUf/aXrhx+y7f4t8IH475cC833wPU7G3fSwAY575IxpmGzQJHAr1qGnzc9eDCuqAgOOhR3+XW4o0+wB+OMMQ1Wgw8UvnA7/p3/BK2Hz5YHD8b17Yc7eSBu/4OiLp4xxkSuQQeK+PxX8M89Dtu2BA/GXTQC19cejDPGmEQNOlC4tnvDkUfjTjnTHowzxpgKpBQoRGQgMJZghbvxqjo66fMmwJNAbyAfGKKqq8PPbgWuAEqAa1R1logcCjybkMVBwO9U9QERuQO4Cvg6/Ow2VZ2+e6dXOdetF65br9rI2hhj6o0qA4WIZAEPAacDucBCEZkSLmda6gpgg6p2EZGhwL3AEBHpRrAsanegPfCqiByiqp8ARyXkvxZ4ISG/Mar6l+qfnjHGmOpK5THiPsByVV2pqkXAJGBwUprBwITw/WSgv4i4cP8kVS1U1VXA8jC/RP2BFar62e6ehDHGmNqTStdTB+DzhO1c4NiK0qhqsYhsBNqG+99OOjb5QYShwDNJ+0aKyKXAIuBGVd2QXCgRGQ4MD7+TnJzMfjo6Ozs748+hJll97GR1UZbVR1l1UR+pBIry7vD6FNNUeqyINAYGAbcmfP4w8Icw3R+AvwI/S85EVccB40rzzPTV4WzB+LKsPnayuijL6qOs6tRH+/btU0qXSqDIBTolbHcE8ipIkysi2UAroCCFY88EFqvqV6U7Et+LyCPAyymU0RhjTC1J5R7FQqCriHQOWwBDgSlJaaYAl4XvLwDmqKoP9w8VkSYi0hnoCryTcNyFJHU7iUi7hM3zgP+lejLGGGNqXpWBQlWLgZHALOCjYJcuEZG7RGRQmOxRoK2ILAduAG4Jj10CKLAUmAn8UlVLAESkOcFIqn8lfeV9IvKhiHwAnApcX81zNMYYUw3O++TbDRnJ5+Ul94ZlFut3LcvqYyeri7KsPsqqgXsUVT5pXG8CRdQFMMaYDFVloKgvy7G5TH+JyLtRlyGdXlYfVhdWH3VWH1WqL4HCGGNMLbFAYYwxplIWKNLHuKqTNChWHztZXZRl9VFWrddHfbmZbYwxppZYi8IYY0ylGvTCRbVJRDoRrNGxHxAHxqnqWBFpQ7AWx4HAakBUdUM42+5Y4EfAVmCYqi4O87oMuD3M+m5VnVCX51KTwmnlFwFrVfXs8In9SUAbYDFwiaoW7eoaJ3V/JtUnIq2B8UAPgiHePwM+oQFeHyJyPXAlQT18CFwOtKOBXBsi8hhwNrBOVXuE+2rsb4WI9AaeAJoB04Frw9kzUmItitpTTDDz7eFAX+CX4foctwCzVbUrMDvchmDeq67hazjB5IilF8vvCWbs7QP8XkT2qssTqWHXEjzhX+pegvVHugIbCP6RQ8IaJ8CYMB1Ja5wMBP4vDD6ZaCwwU1UPA3oS1EuDuz5EpANwDXB0+Ecyi+D/cUO6Np4gKHOimrwWHg7Tlh6X/F2VskBRS1T1i9Ior6qbCf4IdKDs2h0TgHPD94OBJ1XVq+rbQOtw3qszgH+rakE43fq/2cX/yelCRDoCZxH8iib8ZdSPYA0T+H59lNbTrq5xkvZEpCVwEsH0N6hqkap+Q8O9PrKBZuGkos2BL2hA14aq/odgItVENXIthJ+1VNW3wlbEkwl5pcQCRR0QkQOBXsACYF9V/QKCYALsEyYrb92PDpXsz0QPAL8m6IqDYM2Sb8L5xKDsuZVZ4wRIXOOkPtTHQQTL/T4uIu+JyHgRaUEDvD5UdS3wF2ANQYDYCLxLw702StXUtdAhfJ+8P2UWKGqZiOwBPA9cp6qbKkla3hOSVa7pkSlEpLT/9d2E3ZWdW72uD4Jf0D8AHlbVXsAWdnYtlKfe1kfYPTIY6EywZHILgu6VZA3l2qjKrp5/tevFAkUtEpFGBEHiaVUtnSX3q9Kp1MP/rgv3V7R2RyrrgWSCE4BBIrKa4AZlP4IWRuuwuwHKntt3572La5xkilwgV1UXhNuTCQJHQ7w+TgNWqerXqrqDYEbp42m410apmroWcsP3yftTZoGiloR9po8CH6nq/QkfJa7dcRnwUsL+S0XEiUhfYGPY3JwFDBCRvcJfXgPCfRlFVW9V1Y6qeiDBDcc5qnoxMJdgDRP4fn3szhonGUFVvwQ+F5FDw139Cabjb4jXxxqgr4g0D//dlNZFg7w2EtTItRB+tllE+ob1e2lCXimxQFF7TgAuAfqJyPvh60fAaOB0EVlGsB7H6DD9dGAlwQ24R4BfAKhqAcGSsAvD113hvvriN8AN4VombQlv7rIba5xkoF8BT4drrxwF/JEGeH2ErarJBENgPyT4uzSOBnRtiMgzwFvAoSKSKyJXULPXwtUEg0iWAyuAGbtSPnsy2xhjTKWsRWGMMaZSFiiMMcZUygKFMcaYSlmgMMYYUykLFMYYYyplgcIYY0ylLFAYY4yplAUKY4wxlfr/dSUAhwZdRWsAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Время выполнения алгоритма MyKmeans++\n",
    "\n",
    "skm = MyKmeans(k=5, init='k-means')\n",
    "kmpptimes = np.zeros(10)\n",
    "iters = 10\n",
    "size_grid = np.linspace(1000, 10000, 10, dtype=int)\n",
    "for i, size in enumerate(size_grid):\n",
    "    X, y = make_blobs(n_samples=size, n_features=2, centers=5, cluster_std=1, \n",
    "           center_box=(-10.0, 10.0), shuffle=False, random_state=1337)\n",
    "    for j in range(iters):\n",
    "        t1 = time.time()\n",
    "        skm.fit(X)\n",
    "        kmpptimes[i] += time.time() - t1\n",
    "    kmpptimes[i] /= iters\n",
    "plt.plot(size_grid, kmpptimes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mbkm quitting on iteration 9\n",
      "CPU times: user 13 ms, sys: 7.27 ms, total: 20.3 ms\n",
      "Wall time: 18.7 ms\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<__main__.MiniBatchKMeans at 0x1a21036d10>"
      ]
     },
     "execution_count": 101,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X, y = make_blobs(n_samples=10000, n_features=2, centers=5, cluster_std=1,\n",
    "               center_box=(-10.0, 10.0), shuffle=False, random_state=1337)\n",
    "model = MiniBatchKMeans(k=5, init='random', batch_size=1000, epsilon=1e-1)\n",
    "%time model.fit(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1a221a2b50>]"
      ]
     },
     "execution_count": 115,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Время выполнения алгоритма MyKmeans с MiniBatchMyKmeans\n",
    "\n",
    "mbkmtimes = np.zeros(10)\n",
    "iters = 10\n",
    "size_grid = np.linspace(1000, 10000, 10, dtype=int)\n",
    "for i, size in enumerate(size_grid):\n",
    "    X, y = make_blobs(n_samples=size, n_features=2, centers=5, cluster_std=1, \n",
    "           center_box=(-10.0, 10.0), shuffle=False, random_state=1337)\n",
    "    skm = MiniBatchKMeans(k=5, init='random', batch_size=100, max_iter=300, epsilon=0.3)\n",
    "    for j in range(iters):\n",
    "        t1 = time.time()\n",
    "        skm.fit(X)\n",
    "        mbkmtimes[i] += time.time() - t1\n",
    "    mbkmtimes[i] /= iters\n",
    "plt.plot(size_grid, mbkmtimes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1a2221d750>]"
      ]
     },
     "execution_count": 116,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAD8CAYAAACb4nSYAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzt3XmYFNXV+PHv7RkY9m3GFVBRQAVEAUHcF0QWEVzgghqXSEJi9E3iksTkzWKMSTR5E+P700QJJq8YFa+IisoiAoLIgMiAyuICyDIssso2MMDU+f1RPdrTDkzDVHf1cj7Pw5PpqltVp66d09Wnq+41IoJSSqncEAk7AKWUUqmjSV8ppXKIJn2llMohmvSVUiqHaNJXSqkcoklfKaVyiCZ9pZTKIfmJNLLW9gUeBfKAUc65h+LWXwT8DegMDHPOjY1Zdwvwy+jLB51zTwcRuFJKqcNX45W+tTYPeBzoB3QArrfWdohrthq4FXgubtsWwG+Ac4AewG+stc1rH7ZSSqkjkciVfg9gmXNuBYC1dgwwCFhS2cA5tzK6zovbtg8wxTm3Nbp+CtAXeP4Qx9NHhJVS6siYmhokkvRbAmtiXpfiX7knorptW8Y3staOAEYAOOfYt29fgrtPT/n5+Rw4cCDsMNKG9kdV2h9f076oqjb9Ubdu3cSOkUCb6j45Er0aT2hb59xIYGTl+s2bNye4+/RUVFREpp9DkLQ/qtL++Jr2RVW16Y/jjz8+oXaJ3L1TCrSOed0KWJdgHLXZVimlVMASudKfB7Sz1rYB1gLDgBsS3P9k4A8xP95eAfz8sKNUSikViBqv9J1zB4A78RP4Un+RW2ytfcBaOxDAWtvdWlsKDAGetNYujm67Ffgd/gfHPOCByh91lVJKpZ5Jw/H0Zd26zK4AaZ2yKu2PqrQ/vqZ9UVUANf0a797RJ3KVUiqHaNJXSqkcoklfKRUK2bqZPTMmhx1GztGkr5QKhbw+hh1/+y2yqCTsUHKKJn2lVMqJV4EsnAuA9+K/kIqKkCPKHZr0lVKp99kS2LmdehddAetWI++8GXZEOUOTvlIq5WT+bKhblya3/wzad0TGP4eU7Q47rJygSV8plVLieciCOdCxK6ZefSJ2OOzagUx4MezQcoImfaVUan3+KXy5BdP1XADMiW0xPS9Fpo5HNm0IObjsp0lfKZVSUlIMefmYzt2/WmauuQkiechLOrFesmnSV0qljIggJbPh9DMxDRp9tdw0L8T0uRaZ/y7y2ZJD7EHVliZ9pVTqrFkBm7/4qrQTy/S5BpoV4r0wCvHiJ+FTQdGkr5RKGSkpBhPBnNXzG+tMQT3MtTfDqmXIezNCiC43aNJXSqWMlBRD+46Yxk2qXW/OuRhObIuMewYpL09xdLlBk75SKiVk/RpYvwbT7byDtjGRiH8L57bNyJsvpzC63JHIzFlYa/sCjwJ5wCjn3ENx6wuA0UA3YAsw1Dm30lpbF3gSOBvwgB85594OLnylVKaQ+bMBMF2+WdqJZdp3hG7nIZNeQi7sjWlWmIrwckaNV/rW2jzgcaAf0AG43lrbIa7ZcGCbc64t8AjwcHT5dwGcc2cAvYG/WGv124VSOUgWFMMppyWUxCPX3QpeBfLyf5IfWI5JJAH3AJY551Y45/YBY4BBcW0GAZU32I4FellrDf6HxFQA59xG4Ev8q36lVA6RTRtg9QpM14OXdmKZo47F9LoKKZ6GrFqe5OjSg/fcE5RNGJv04yRS3mkJrIl5XQqcc7A2zrkD1trtQCHwATDIWjsGaI1f/mkNvBe7sbV2BDAiuj1FRUWHfyZpJD8/P+PPIUjaH1XlYn/snvUmu4DCXv3Jizn3Q/WFd9PtbC6eTt7Lo2n+u8cwpsaZADPWnrcnsWP6BKTo6KS/NxJJ+tX1dPzEugdr8y/gdOB9YBUwGzgQ39A5NxIYWbldps+ZqfN+VqX9UVUu9kfFO1PghFPYllcXYs69xr64ahj7n32CzVNer/be/mwg60vxnvgTtOtAg6G31XaO3BolkvRL8a/OK7UC4mcur2xTaq3NB5oCW51zAtxV2chaOxv4LKHIlFJZQbZuhhWfYK7+1mFvay7sg0yfgDf230Q6n43Jr5OECMMj5eV4Tz4MdeoS+e5PMHkJ3VtTK4nU9OcB7ay1baJ34wwDxse1GQ/cEv17MDDNOSfW2gbW2oYA1trewAHnnD5jrVQOkYVzAA55q+bBmLw8IkNug00bkGlvBB1a6GTMSFi3msjwuzHNU3OXUo1J3zl3ALgTmAws9Re5xdbaB6y1A6PNngIKrbXLgLuB+6LLjwZKrLVLgZ8BNwV9Akqp9CYlxXBca8yxrY5oe9OpK3Tqirz+ArJzR8DRhccrno7MmoLpN8Q/xxQxIvHl+dDJunXx1aPMkos120PR/qgql/pDdm7Hu+cWzJVDiAy68RvrE+0LWbca77c/xFzcj8gN30tGqCkl69fgPXg3nNSWyN0PYvLygNq9N6I1/Rp/7dZ75pVSSSML5oB4Cd+qeTDm+BMwF/VFZkz0n+zNYFJejvfEw1BQj8h37/0q4aeKJn2lVNJIyWw46lhodVKt92UGXg8F9fFe/HftAwuRPP8ErF/j1/FDeNpYk75SKimkbBd8/CGm63mB3GNvGjfFXGnho/eRxQsCiDD1vNlTkXenYvoPwXTsEkoMmvSVUkkhH8yDiopA7683lw2Ao47Fe/FfSEVFYPtNBVm3Gnn2CWjfCXPV9aHFoUlfKZUUUjIbmhfBSe0C26epU8cfl2ftKmTWlMD2m2xSvjfUOn4sTfpKqcDJ3j2weAGm67mYSMBppuu50L4j8uqzyJ6yYPedJPLck7ChlMh37sY0axFqLJr0lVLBWzQf9u9LytAJxhh/zP1dO5AJLwa+/6B5705FZk/FXGkxHcKp48fSpK+UCpyUFEPjptD29KTs35zYFtPzUuStV/0RPNOUrF2NPPcPOPUMzFXDwg4H0KSvlAqY7N+HfPg+pktPTCR5tWtzzU0QyUPGjU7aMWpDyvf64+oU1CfynXuS2heHQ5O+UipYixdA+Z5aP5BVE9O8ENPnWuT9Wciy9BvSS559wq/jf/fe0Ov4sTTpK6UCJSWzoUEjOPWMpB/L9LkGmhXivfAU4nlJP16ivHenIsXTMAOGYk4/M+xwqtCkr5QKjBw4gHzwHubMHpj85A8TbArq+WWelZ8h781M+vESIWtXfV3HHzA07HC+QZO+Uio4n3wEZbtTOuGJ6XkJnNgWGTcaKS9P2XGrI3v3+Pfj12vgl3XSpI4fS5O+UiowUjIbCupDCocYMJGIfwvnts3IlJdTdtx4IuLX8b9Y6/9w27R5aLEciiZ9pVQgxKtAFszBdD4bU6duSo9t2neEruchE19CvtyS0mNXknffQuZMxwwYlnZ1/FgJFd2stX2BR4E8YJRz7qG49QXAaPyJz7cAQ51zK621dYBRQNfosUY75/4YYPxKqXSxbCns3B7aXLaRwbfiffge8sp/MLf+KKXHltKV/lO3p5+JGWBTeuzDVeOVvrU2D3gc6Ad0AK631naIazYc2Oacaws8AjwcXT4EKHDOnYH/gfA9a+1JAcWulEojUlIMdepCp26hHN8cdSym11XI7GnIquUpO67s3YP35J+gQUN/mIU0rOPHSqS80wNY5pxb4ZzbB4wBBsW1GQQ8Hf17LNDLWmsAARpGJ0uvD+wDsme+M6UUAOJ5ftLv2AVTr35ocZj+Fho29kfhTMGsgH4d/x/wxTq/jt8kPev4sRJJ+i2B2KlqSqPLqm0TnVN3O1CI/wGwG1gPrAb+xzm3tZYxK6XSzcrPYNvmpD+QVRPToCFm0A3+XUQL5yb9eDJrCjLnbcxVwzCndU768YKQSE2/utkP4j9CD9amB1ABHA80B96x1r7lnFsR29BaOwIYAeCco6ioKIGw0ld+fn7Gn0OQtD+qysb+2PnGC5Tl51N0WV8iDRsnvF0y+kKuuYEtMyfDy6MpvKQPpk6dQPdfaf+q5WwdM5K6Z3an2c23BzJccireG4kk/VKgdczrVkD8zOWVbUqjpZymwFbgBmCSc24/sNFa+y5wNlAl6TvnRgIjoy8l0yeNzqWJrxOh/VFVtvWHiOC9OxVO68zWPeWwJ/F75ZPVF3LtzXiP/pZNY0cT6R1fjQ5g/3vL8B66D+o35MDNd7Jl27ZA9hvAxOg1SqS8Mw9oZ61tY62tCwwDxse1GQ/cEv17MDDNOSf4JZ3LrLXGWtsQ6Al8nFBkSqnMULoSNm0IvbQTy3TqBp26Iq+PQXYF+zOiiCDP/AO+WJ8xdfxYNSb9aI3+TmAysNRf5BZbax+w1g6MNnsKKLTWLgPuBu6LLn8caAQswv/w+Ldz7sOAz0EpFSIpmQ0mgjnrnLBDqSIy5DbYuwcZ/3yg+5VZU5D3ZmAGZk4dP5ZJxS/ch0nWrYuvHmWWbPv6XlvaH1VlW39U/PoOaNKMvHt/f9jbJrsvvGf/gcycTOT+xzDHtar1/qT0c7w//ATadSDyo98EfntmAOWdGmeg1ydylVJHTNaXwvo1oT2QVRMz8AYoqI/34r9qvS/ZW4b3xJ+gQSMiw9P/fvyD0aSvlDpiUjIbANMlTZN+46aYKy189D6yZMER78ev4/8dNq73B1Jr0izAKFNLk75S6ohJSTGcfCqmeWHYoRyUuWwAHHUsnvsX4lUc0T7knTeR92ZiBt2AObVTwBGmliZ9pdQRkU0bYPXytLprpzqmTh0i190Ka1chs6Yc9vay5nPk+ZHQ4SxMv8HBB5himvSVUkdEFhQDpG09v4qu50L7jsgrzyJ7yhLezK/jPwyNGkfr+JmfMjP/DJRSoZCSYjjhZMxRx4YdSo2MMf6Y+zu3IxNfTGibr+r4mzZkfB0/liZ9pdRhky+3wPKP0760E8uc2BZz7qXIlPHI5i9qbC8zJ39dx2+f2XX8WJr0lVKHTRbMATKktBPDXHMzRCLIS08fsp2sXoGM+ac/amgW1PFjadJXSh02mT8bjmuNOa51zY3TiGleiOlzLfL+LGTZ0mrbyJ4yf3z8LKrjx8qus1FKJZ3s3A6fLs64q/xKps810KwFnnsK8bwq6/w6/uPROv5PMI2bhhRl8mjSV0odFlk4F8TLqHp+LFNQzy/zfP4p8t7MKutkxiRk3juYq2/0593NQpr0lVKHRUqKoegYaN0m7FCOmOl5CZzYFhk3Gin3h4KW1cuRF0ZBp66YvteFG2ASadJXSiVMynbB0g8wXc/DmBrH9kpbJhLxb+HcthmZ8kpMHb8Jkdvuyro6fqxEJlFRSikA5MN5UHEgY+v5sUz7jtD1PGTSS8iKT2DzF0Tu/UNW1vFjZe/HmVIqcFJSDM0KoU37sEMJROS6W6DiAHz0PubqmzDtOoQdUtIldKVvre0LPArkAaOccw/FrS8ARgPdgC3AUOfcSmvtjcBPYpp2Bro65xYGEbxSKnWkfC8sLsFccEXWlD/M0cdh7HBYt8a/qycH1Phfzlqbhz8DVj+gA3C9tTb+43A4sM051xZ4BHgYwDn3rHPuLOfcWcBNwEpN+EplqEXzYd++rCjtxIpceiWRG7+fNR9kNUnkLHsAy5xzK5xz+4AxQPxMw4OAykfcxgK9rLXxv/JcDwQ7b5lSKmVk/mxo3BRyoASSzRJJ+i2BNTGvS6PLqm0TnVN3OxA/wPZQNOkrlZFk/z7kw/cxZ52TsTNGKV8iNf3q7suKn1j3kG2stecAZc65RdUdwFo7AhgB4JyjqKgogbDSV35+fsafQ5C0P6rKxP4on/cuX5bvoemlfSkIMPZM7ItkSkV/JJL0S4HYATZaAfEzl1e2KbXW5gNNga0x64dxiKt859xIYGT0pWT6pNHZNvF1bWl/VJWJ/eG9PQkaNGTHcSdiAow9E/simQKYGL1GiST9eUA7a20bYC1+Ar8hrs144BagGBgMTHPOCYC1NgIMAS5KKCKlVFqRAweQhXMxnXtg8uuEHY6qpRpr+tEa/Z3AZGCpv8gtttY+YK0dGG32FFBorV0G3A3cF7OLi4BS59yKYENXSqXEpx9B2S5Mt+y6aydXGZH48nzoZN26+OpRZtGvrFVpf1SVaf3h/efvyJy3ifz1GUzdgkD3nWl9kWwBlHdqHBsjN25MVUodEfEqkAVzMGecHXjCV+HQpK+UOrhlH8OOL/2JxVVW0KSvlDooKZkN+XUwZ3QLOxQVEE36SqlqiQiyoNifJ7Zeg7DDUQHRpK+Uqt7KZbB1c8bOkKWqp0lfKVUtKZkNeXmYM3uEHYoKkCZ9pdQ3iIif9E/tjGnYKOxwVIA06SulvmntSti4Xh/IykKa9JVS3yDzi8FEMGf1DDsUFTBN+kqpb5AFxdCuA6ZJs7BDUQHTpK+UqkI2rIW1q/SunSylSV8pVYWUzAbAdNHSTjbSpK+UqkJKiqFNe0wLndwkG2nSV0p9RbZshFXLMN20tJOtNOkrpb4iJcUAmC56q2a2SmTmLKy1fYFHgTxglHPuobj1BcBooBuwBRjqnFsZXdcZeBJoAnhAd+fc3qBOQCkVHCmZDa3aYI4+LuxQVJLUeKVvrc0DHgf6AR2A6621HeKaDQe2OefaAo8AD0e3zQf+A3zfOdcRuATYH1j0SqnAyJdbYfnH+kBWlkvkSr8HsKxyukNr7RhgELAkps0g4P7o32OBx6y1BrgC+NA59wGAc25LQHErpQImC+aAiN6qmeUSqem3BNbEvC6NLqu2TXRO3e1AIdAeEGvtZGttibX2p7UPWSmVDLKgGI5thTn+hLBDUUmUyJV+dXMuxk+se7A2+cAFQHegDJhqrZ3vnJsa29BaOwIYAeCco6gos28Vy8/Pz/hzCJL2R1Xp2B/eju1s+mQRDa/9Fo1SGFs69kWYUtEfiST9UqB1zOtWQPzM5ZVtSqN1/KbA1ujyGc65zQDW2glAV6BK0nfOjQRGRl9Kpk+UrJM9V6X9UVU69oc3awp4Few57Sz2pjC2dOyLMAUwMXqNEkn684B21to2wFpgGHBDXJvxwC1AMTAYmOacE2vtZOCn1toGwD7gYvwfepVSaURKiqHwaDjh5LBDUUlWY00/WqO/E5gMLPUXucXW2gestQOjzZ4CCq21y4C7gfui224D/or/wbEQKHHOvRH8aah4smsHUroy7DBUBpCy3bB0IabbeRhTXaVWZRMjEl+eD52sWxdfPcos6fCVteKxB2HJQiIP/RPTpHmosaRDf6STdOsPb+4MZNRfiNz3J8wpp6X02OnWF2ELoLxT46e2PpGbhaR0JXzwHuzfh7w1PuxwVJqTktnQrAW0aR92KCoFNOlnIZn4EhTUh05dkbcnImW7wg5JpSkp3wuL5mO69MRENB3kAv2vnGVk43pk3juYS/oSueZm2FOGTJ8QdlgqXS0qgX379IGsHKJJP8vI5HGQl4e5fBDmhJOhUzfkrfFIeXnYoak0JCXF0KgJtOsYdigqRTTpZxH5cgsyeyrm/F6YZi0AiPQfArt2ILOmhBydSjeyfz/y0Ty/tJOXF3Y4KkU06WcRmfIqeB6mz7VfLTPtOkC7Dsib45ADOtadirF0Iewp02GUc4wm/Swhu3ciMyZhul+IOerYKusi/YbA1s3I3BkhRafSkZTMhvoN4fTOYYeiUkiTfpaQqa9D+V5Mv8HfXNmpK7Rug0x6CfEqUh+cSjtSUYEsfA9zZndMfp2ww1EppEk/C8jePcjU1+DMHpiWJ35jvTHGr+1vWAsL5oQQoUo7ny6C3Tv1rp0cpEk/C8jMyVC2i0h1V/mVup4Lx7TEm/AiafgUtkoxKZkNdQugY5ewQ1Eppkk/w8n+/cibr8BpnQ/5CL2J5GH6XgurV8DiBSmMUKUb8Tx/wpQzumHqFoQdjkoxTfoZToqnwvath77KjzI9L4HmRXgTX0x+YCotyarlyP/9L2zfpqWdHJXQxOgqPUlFBTJpHJzUDk4/s8b2Jr8O5oqrkRdGIcuWYNrGT3WsspGU70Xem+mXAVd+BnXrYi7qq7dq5ihN+hlM3p8FmzYQGXJbwkPimguvQN5weBPGkvfDXyc5QhUmWbsKmTEJmTMd9pTB8Sdgrh+B6XkJpkGjsMNTIdGkn6FEBJk4Fo5rDWf2SHg7U1APc/lA5JX/IKtX+EM1qKwh+/ch899FZkyCZUshvw7m7PMxF/WFtqfrePlKk37G+vB9WLsKc9tdhz06orm0v3/P/qSXMCN+kqQAVSrJhrXIzEnI7GmweyccfTxmyLcx5/bCNG4SdngqjSSU9K21fYFHgTxglHPuobj1BcBooBuwBRjqnFtprT0Jf7atT6JN5zjnvh9Q7DlLRPwfYwuPxnS/8LC3Nw0aYS7pj0x+GRl0I+aYxObWVOlFDuxHFsxFZk6Cjz/0B9o7qyfm4r7+3Vx6Va+qUWPSt9bmAY8DvfEnOp9nrR3vnFsS02w4sM0519ZaOwx4GBgaXbfcOXdWwHHntk8Xw/KPMTd8H5N/ZF/WTO+ByNTXkMnjMDffGXCAKplk0wbknTf9QfR2bvc//K+5CXP+5Zim4c6SptJfIhmjB7DMObcCwFo7BhgExCb9QcD90b/HAo9Za/UyI0m8CS9Ck2aY83sd8T5Mk+aY8y/3k8eAYZgWRQFGqIImFRXw0Ty8GZOiz1kY6Hw2kYv7QcezMBEdJVMlJpGk3xJYE/O6FDjnYG2ccwestduBwui6NtbaBcAO4JfOuXfiD2CtHQGMiG5PUVFmJ6D8/PykncP+5R+zdckCGt10Ow2Pb1mrfVVcP5zNMydTb9ZkGt/2o4Ai/KZk9kcmOpz+qNi8kT1vjWfPW6/hbdlEpEUR9e23qX/5VeQVHZPkSJNP3xtVpaI/Ekn61V2xxz/Hf7A264ETnHNbrLXdgFestR2dcztiGzrnRgIjK7fL9ImSkznZc8Vzo6B+Q8q6X8ye2h4jUgfT4yLKJr/C3kuvStoPfjr5dVU19Yd4FbB4Id7MSfDBPECgYxciw74LZ3Rnb14eewGyoE/1vVFVABOj1yiRpF8KtI553QpYd5A2pdbafKApsNU5J0A5gHNuvrV2OdAeeD+h6FQVsr4UFhRj+g/B1G8QyD5Nv+uQOdORaa9hBt0YyD7VkZEd25BZb/kPUW3ZCI2bYvpeg7mwzzeGy1bqSCWS9OcB7ay1bYC1wDDghrg244FbgGJgMDDNOSfW2qPwk3+FtfZkoB2wIrDoc4xMegnq1MH0uiqwfZrjT4AuPZFpryN9rsHUC+bDRCVGRODjD/2HqBbOgYoKOPUMzHW3Yrqco8Meq8DVmPSjNfo7gcn4t2z+yzm32Fr7APC+c2488BTwjLV2GbAV/4MB4CLgAWvtAaAC+L5zbmsyTiTbyZaNyNy3MZf0xzRuGui+I/2G4C2Y40/CEjPrlkoe2bUDmT0VmTEZNq6Dho0xlw3AXNQHc2yrsMNTWcyk4TC7sm5dfPUosySjTuk99yQyczKRPzyJaXFUoPsGqPjrr2DdaiJ//CemTt1A9611W5+IwPKl1J0znb3vToMD+/2nZC/ui+l2fuD9ngn0vVFVADX9Gu+a1CdyM4Ds+BKZNcUfMyUJCR/8CdS9v/wSeXcq5pJ+STlGrpLycuS9Gci0N6D0c8obNMRc2Nsf9KzVSWGHp3KMJv0MIG+NhwP7MX2vS95BTj0D2rRHJo9DLrwCk6f3fdeWbP4CeXsCMustf2iElidibvoBR/W/ji27docdnspRmvTTnJTtRt6egOl6HubY2t2XfyiVUyp6j/8emfeOP/a+OmwiAks/wJv2Onw4D4yBLj2JXDoA2nfEGIOpVx806auQaNJPc/L2BNhThulf8yQptda5O7Q8EZk4Fulx0WEP5JbLZG8ZUjzdL+FsKIVGTTD9Bvv1+iSV5JQ6Epr005iUl/ulnU5dMSeckvTjmUgE0/c65Km/wofvwVk9k37MTCcb1volnNlT/THrT2yL+faPMd0vyMkfZlX606SfxuRdf0CtSL8hKTum6X4h8uqzeBPGEjnzHB2psRriebBovl/CWbwA8vL9MesvGwBt2mufqbSmST9NyYH9yORx0LYDpn3HlB3X5OX5V/v/+bs/XG8C0zDmCinb5T8x+/YE2LQBmrbADLzBv7deR7dUGUKTfpqSuTNh62Yi3/pByo9tzrsMee15vIljydOk7087OO0Nf9rBfeX+vfXX3ITpcu4RD22tVFj0HZuGxKtAJo2FVm2gU7eUH9/UqYvpfTUy9t/I559i2rRPeQxhk4oK+GAu3rQ34JOPoE5dTI+LMJddmZLfV5RKFk366WjBXNiwFjPip6HVh83FfZAJL/oTqN/xi1BiCIPs3IG8MxmZMRG2bvYnKLnuFswFvTGNdNpBlfk06acZEfEnSTn6eEy3c0OLw9RrgLlsAPL6GGTtakzLE0KLJRVk1TJk6uvIvHf84RFOP5PIsBFwZnedoERlFU366WbJQli9HHPznaEnG9NrADLlFX8C9eF3hRpLMsiB/cj82cj0N2D5x1BQD3PB5ZhLr/RHH1UqC2nSTzPehBehWSHm3EvDDgXTqAnmwj7ItNeQgddnzZju8uVWZOYkf9z67dvg6OMwQ7+DOa8XpkHDsMNTKqk06acRWbYUPl2EGTo8bcZRN1dcjUx/A3nzZcyNt4cdzhHzR7j82J83oGS2P279GWcTuexK6NBFnz5WOUOTfhrxJo6FRo0xF/YJO5SvmOaF/i2cs97yJ1DPwPvRZceXeE88BJ8tgfoNMZcOwFzaD3N0YtPLKZVNEkr61tq+wKP4k6iMcs49FLe+ABgNdAO2AEOdcytj1p8ALAHud879TzChZxcp/Rw+nIcZdAOmoF7Y4VRh+l7rJ/0pr2IG3xp2OIdFNm3A+9tv4MstmOtH+CWcevXDDkup0NT4ndZamwc8DvQDOgDXW2s7xDUbDmxzzrUFHgEejlv/CDCx9uFmL5n4EhTUx1w6IOxQvsEcfTzm7PORtyciu3eFHU7CZM3neA//DHbtJHL3g0QuG6AJX+W8RAqZPYBlzrkVzrl9wBhgUFybQcDT0b/HAr2stQbAWns1/ry4i4MJOfvIxnXIvFmYS/phGjYKO5xqmf6DoXyPf6dLBpBPFuH9+ecQySPys4cwp5zltR80AAASDElEQVQWdkhKpYVEyjstgTUxr0uBcw7WJjqn7nag0Fq7B/gZ0Bu492AHsNaOAEZEt6eoqCjhE0hH+fn5h3UOO9wo9uTnUzj02+Q1L0xiZLVQVMS2buexf9rrFA677bCumA+3P2prb/HbbH/0fvKOOZ7mv3mEvKJjUnbsRKS6P9KZ9kVVqeiPRJJ+dY+Exk+se7A2vwUecc7tstYe9ADOuZHAyMrtMn3OzMOZ51K2bcGbPgFzQW+2VQik8bnL5YOQ+bPZ9MpzRC6P/7J3cKmcB9WbMQl59gk4uT3yX79iG3lp16c6L+zXtC+qCmCO3BolUt4pBVrHvG4FxM9c/lUba20+0BTYiv+N4E/W2pXAj4FfWGvvTCiyHCFTXgHPw1xxTdih1Mi0PR3ad0Imv4Ic2B92OFWICN5rY/zRQTt1JXLX7zANG4cdllJpJ5Er/XlAO2ttG2AtMAy4Ia7NeOAWoBgYDExzzglwYWUDa+39wC7n3GMBxJ0VZNcOZOZkfyCvDHnwKdJvMN6j9yPF0zEXXhF2OEB0gLrnRyJvT8Sce5n/NLOOfqlUtWq80nfOHQDuBCYDS/1FbrG19gFr7cBos6fwa/jLgLuB+5IVcDaRaa9D+V5M3xRMhRiUjl3ghFOQSeMQryLsaJD9+/Ce/LOf8Ptci/n2jzThK3UIRiS+PB86WbcuvnqUWRKpy8neMryffQfadyTvjv9OUWTBkPmz8Z54CDPip0S6X1Bj+2TVbaVsN97f/wCffIQZchuRK64O/BjJoHXsr2lfVBVATb/GYXn12fOQyMzJULaLSL8Musqv1KUnHNsSmfAiYV00yPZteP/zC1i2BDP87oxJ+EqFTZN+CGT/PuTNV+H0MzEnnxp2OIfNn0B9MJR+Dovmp/z4snGd/9DVxvVE7vwVkZ6XpDwGpTKVJv0QyOxpsH1rZl7lR5lzLoYWR+FNGJvS48qq5XgP/Qz27CZyz4OYTl1TenylMp0m/RSTigp/wvM27eG0zmGHc8RMfr5/m+myJcinqXnYWpZ+gPfnX0CdukR+9nBOTuOoVG1p0k8xeX8WbNpApN/g0KZCDIq5oDc0boo38cWkH8ubNwvvf38LRUcTue9PmGNbJf2YSmUjTfopJJ6HTBwLx7WGM3uEHU6tmYICTK+rYFEJsnp50o7jTX8D+eef4aT2RH7yR0y6DlWhVAbQpJ9KH70Pa1dh+g/Omkk7zKX9oX4DJAm1fRHBe+U/yHNPQufuRO76bdoOSKdUpsiOzJMBvprwvPBoTPeLwg4nMKZBI8wl/ZGS2ciG0sD2KxUVyDOPI284zAW9idz+c0zdgsD2r1Su0qSfKp8ughWfYPpei8kLd8LzoJnLB0J+HWTSuED2J/vK8Z54GHnnTUx/6w+rkGV9plRYNOmniDfhRWjSDHP+5WGHEjjTpBnmgt7InOnIlk212peU7fJnuvpgLmbYCCLXfCvjf/BWKp1o0k8BWfkZLFmI6T0IU6du2OEkhenjjxIqU1454n3Il1vw/vRzWPEp5jv3EOmVfrOIKZXpNOmngDdxLDRoiLm4X9ihJI0pPBrT42LkncnIzu2Hvb1sWOs/dLV5I5Ef/ppIj+z53UOpdKJJP8lk/RooKcZceiWmfoOww0kq0+862L8feeu1w9pOPv/MH1ZhXzmRex/EdDgrSREqpTTpJ5lMfAnqFmB6Day5cYYzx7WGLuci099A9pQltI0sXoD3l/+Ggnr+U7YntUtukErlOE36SSSbv0Dmvo25qA+mcZOww0mJSP/BsGc38vbEGtt6c2fg/b8H4Kjj/Kdsj0lsujel1JFLaLYJa21f4FEgDxjlnHsobn0BMBroBmwBhjrnVlpre/D13LcGuN8593JQwac7efNlMBFM79wZ9tec2BY6dEGmvIL0GnDQe+u9t8YjL4yC9p2I3PHfmAYNUxypUrmpxit9a20e8DjQD+gAXG+t7RDXbDiwzTnXFngEeDi6fBFwtnPuLKAv8GR0Dt2sJzu2IbPewpx7KaZFcme3TzeR/kNg53bk3be+sU5E8MY97Sf8rucS+fH9mvCVSqFEEnAPYJlzbgWAtXYMMAhYEtNmEHB/9O+xwGPWWuOciy3s1gPSbpquZJG3xsOB/Zg+14YdSuq17winnIZMfhm5sM9Xi/2nbB9D3p2Kuagv5sbvYSL60JVSqZRITb8lsCbmdWl0WbVtonPqbgcKAay151hrFwMfAd+Prs9q3u6dyPQJmG7nY46N76rsZ4wh0m8IbNmIvDcTACkvx/v7H/yEf9UwzLdu14SvVAgSudKv7nHI+Cv2g7Zxzs0FOlprTweettZOdM7tjW1orR0BjIi2p6gos8she8Y9A3v30PyG71Anw8/lSMllfdn6+vPIlFcwvfqT99gDeJ8sovH37qVB3xz89hMjPz8/49/jQdG+qCoV/ZFI0i8FWse8bgXEz1xe2aY0WrNvCmyNbeCcW2qt3Q10At6PWzeSr3/wlUyeKFnKy5FXn4dO3djeuAVk8LnUlnf5IGTUX9h4x1C83TuJfO+nlHU7n7Ic7hPQycBjaV9UFcDE6DVKJOnPA9pZa9sAa4FhwA1xbcYDtwDFwGBgmnNOotuscc4dsNaeCJwKrEwosgwknoe88E9kx5cZPRViUMzZFyDjn0N2bifyo/sxGTxTmFLZosakH03YdwKT8W/Z/JdzbrG19gHgfefceOAp4Blr7TL8K/xh0c0vAO6z1u4HPOAHzrms/FgXEeT5J5F33qTh4FvY275j2CGFzuTlEbnn97Ro0ZxtaP1eqXRgRNLuhhpZty6+epTe/IQ/Epn+BqbvdRw14m62bNkSdlhpQ7/CV6X98TXti6oCKO/UOCStPpFbSyKCvDDKT/hXXI259mYdClgplbY06deCiCBj/41MfQ1z+UDM4G9rwldKpTVN+kdIRJBxo5E3X/FH0LTDNeErpdKeJv0jICLIK88ik17CXNIPc/0ITfhKqYygSf8IyGvPIxMc5sIrMNd/TxO+UipjaNI/TN7rY5DXxmDOvxzzrR9gItqFSqnMoRnrMHgTXkRefQ5z7mWYm+/UhK+UyjiatRLkTR6HvPwM5pyLMbf+lyZ8pVRG0syVAG/Kq8jY/8N0vxDz7R/r6JBKqYylSb8G3tTXEPeUP0zy8LsxeZrwlVKZS5P+IXjT30DG/BO6nov5zj2a8JVSGU+T/kF4MyYhzz0JZ51D5Lv3YvJzYpZHpVSW06RfDe+dN5H//B06dyfyvZ9i8uuEHZJSSgVCk34c792pyDOPQ6duRL5/nyZ8pVRW0aQfwyuejjz9v3D6WUR+8HNMHU34SqnsklCh2lrbF3gUfxKVUc65h+LWFwCjgW7AFmCoc26ltbY38BBQF9gH/MQ5Ny3A+APjzZ2B/PtROK0zkTt+galTN+yQlFIqcDVe6Vtr84DHgX5AB+B6a22HuGbDgW3OubbAI8DD0eWbgaucc2fgT6f4TFCBB8mb9w7y1CPQviORO36JqVsQdkhKKZUUiZR3egDLnHMrnHP7gDHAoLg2g4Cno3+PBXpZa41zboFzrnIarMVAvei3grQh82cjo/4CbU8j8l+/whSkVXhKKRWoRJJ+S2BNzOvS6LJq2zjnDgDbgcK4NtcBC5xz5UcWavCkpBjvn3+Gk08l8sNfYwrqhR2SUkolVSI1/erGDY6fWPeQbay1HfFLPldUdwBr7QhgBIBzjqKiogTCqp29773D9pF/ok7b02n260eINGgY2L7z8/NTcg6ZQvujKu2Pr2lfVJWK/kgk6ZcCrWNetwLiZy6vbFNqrc0HmgJbAay1rYCXgZudc8urO4BzbiQwMvpSkj1Rsnw4D+/vf4QTTqHiB//N1rI9ULYnsP3rZM9VaX9Upf3xNe2LqgKYGL1GiST9eUA7a20bYC0wDLghrs14/B9qi4HBwDTnnFhrmwFvAD93zr2bYOxJJYtK8P7xR2h1EpEf348J8ApfKaXSXY01/WiN/k5gMrDUX+QWW2sfsNYOjDZ7Cii01i4D7gbuiy6/E2gL/MpauzD67+jAzyJBsmQh3uO/h+NPIHLXbzENGoUVilJKhcKIxJfnQyfr1sVXjwLY6dIP8P7f7+CY44nc8yCmUZPAj1FJv7JWpf1RlfbH17QvqgqgvFPj3K058USufLII77HfwdHHEbn7d0lN+Eoplc6yPunLp4vx/t8DUHiMn/AbNw07JKWUCk1WJ31ZthTvfx+A5kVE7n0Q06RZ2CEppVSosjbpy4pP8B69H5q18Gv4TZqHHZJSSoUuK5O+fP4Z3t9+A02a+Qm/WYuwQ1JKqbSQdUlfVi3D+9uvoVETIvf8HtM8fjQIpZTKXVmV9GX1Cry//hrqN/QTfgt9vFsppWJlTdKXtavwHvkV1KtP5N7fYwqPCjskpZRKO9kz23fDxnDCKURuvB1TdEzY0SilVFrKmqRvmrUg764Hwg5DKaXSWtaUd5RSStVMk75SSuUQTfpKKZVDNOkrpVQO0aSvlFI5RJO+UkrlEE36SimVQzTpK6VUDknL6RLDDkAppTJURk6XaDL9n7V2ftgxpNM/7Q/tD+2LlPVHjdIx6SullEoSTfpKKZVDNOknx8iwA0gz2h9VaX98TfuiqqT3Rzr+kKuUUipJ9EpfKaVySNaMp59s1trWwGjgWMADRjrnHrXWtgBeAE4CVgLWObfNWmuAR4H+QBlwq3OuJLqvW4BfRnf9oHPu6VSeS1CstXnA+8Ba59wAa20bYAzQAigBbnLO7bPWFuD3XTdgCzDUObcyuo+fA8OBCuCHzrnJqT+T2rPWNgNGAZ3wbzu+DfiEHHxvWGvvAr6D3w8fAd8GjiNH3hvW2n8BA4CNzrlO0WWB5QlrbTfg/4D6wATgR865hEs2eqWfuAPAPc6504GewB3W2g7AfcBU51w7YGr0NUA/oF303wjgH/DVf/zfAOcAPYDfWGubp/JEAvQjYGnM64eBR6J9sQ3//7BE/3ebc64t8Ei0HdH+GwZ0BPoCf49+kGSiR4FJzrnTgDPx+yXn3hvW2pbAD4GzowkvD/+/cS69N/4PP+ZYQb4X/hFtW7ld/LEOSZN+gpxz6ys/gZ1zO/H/T90SGARUXo09DVwd/XsQMNo5J865OUAza+1xQB9ginNuq3NuGzCFw/yPlg6sta2AK/GvbolesVwGjI02ie+Lyj4aC/SKth8EjHHOlTvnPgeW4b/BM4q1tglwEfAUgHNun3PuS3L0vYFfQahvrc0HGgDryaH3hnNuJrA1bnEg74XouibOueLo1f3omH0lRJP+EbDWngR0AeYCxzjn1oP/wQAcHW3WElgTs1lpdNnBlmeavwE/xS91ARQCXzrnDkRfx57XV+ccXb892j5b+uJkYBPwb2vtAmvtKGttQ3LwveGcWwv8D7AaP9lvB+aTu++NSkG9F1pG/45fnjBN+ofJWtsIeAn4sXNuxyGaVvd0nBxiecaw1lbWK+fHLD7UeWVtX0TlA12BfzjnugC7+frre3Wytj+iJYhBQBvgeKAhfgkjXq68N2pyuOdf637RpH8YrLV18BP+s865cdHFX0S/chH9343R5aVA65jNWwHrDrE8k5wPDLTWrsT/ce4y/Cv/ZtGv9FD1vL465+j6pvhff7OhL8A/j1Ln3Nzo67H4HwK5+N64HPjcObfJObcfGAecR+6+NyoF9V4ojf4dvzxhmvQTFK0zPgUsdc79NWbVeOCW6N+3AK/GLL/ZWmustT2B7dGvdZOBK6y1zaNXRVdEl2UM59zPnXOtnHMn4f/YNs05dyMwHRgcbRbfF5V9NDjaXqLLh1lrC6J3/rQD3kvRaQTGObcBWGOtPTW6qBewhBx8b+CXdXpaaxtE/z9T2Rc5+d6IEch7Ibpup7W2Z7R/b47ZV0I06SfufOAm4DJr7cLov/7AQ0Bva+1nQO/oa/BvpVqB/wPUP4EfADjntgK/A+ZF/z0QXZYNfgbcba1dhl+XfSq6/CmgMLr8bqKlD+fcYsDhJ4VJwB3OuYqURx2M/wKetdZ+CJwF/IEcfG9Ev+2Mxb8t8yP8HDOSHHpvWGufB4qBU621pdba4QT7Xrgd/waKZcByYOLhxKdP5CqlVA7RK32llMohmvSVUiqHaNJXSqkcoklfKaVyiCZ9pZTKIZr0lVIqh2jSV0qpHKJJXymlcsj/BzXVft4kd/46AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Время выполнения алгоритма  MyKmeans++ с MiniBatchMyKmeans\n",
    "\n",
    "mbkmpptimes = np.zeros(10)\n",
    "iters = 10\n",
    "size_grid = np.linspace(1000, 10000, 10, dtype=int)\n",
    "for i, size in enumerate(size_grid):\n",
    "    X, y = make_blobs(n_samples=size, n_features=2, centers=5, cluster_std=1, \n",
    "           center_box=(-10.0, 10.0), shuffle=False, random_state=1337)\n",
    "    skm = MiniBatchKMeans(k=5, init='k-means', batch_size=100, max_iter=300, epsilon=0.3)\n",
    "    for j in range(iters):\n",
    "        t1 = time.time()\n",
    "        skm.fit(X)\n",
    "        mbkmpptimes[i] += time.time() - t1\n",
    "    mbkmpptimes[i] /= iters\n",
    "plt.plot(size_grid, mbkmpptimes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1a23226f10>"
      ]
     },
     "execution_count": 128,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(12, 8))\n",
    "plt.ylabel('time')\n",
    "plt.xlabel('number of samples')\n",
    "plt.plot(size_grid, skmtimes, label='sklearn kmeans')\n",
    "plt.plot(size_grid, kmtimes, label='my kmeans')\n",
    "plt.plot(size_grid, kmpptimes, label='my kmeans++')\n",
    "plt.plot(size_grid, mbkmtimes, label='my minibatch kmeans')\n",
    "plt.plot(size_grid, mbkmpptimes, label='my minibatch kmeans++')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Применение K-means на реальных данных"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Загрузите [данные](https://github.com/brenden17/sklearnlab/blob/master/facebook/snsdata.csv) в которых содержится описание интересов профилей учеников старшей школы США.  (без этого пункта задание не считается выполненным).   \n",
    "ВНИМАНИЕ! Проверка должна быть осуществлена на всех реализациях, иначе реализация не будет зачтена!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>gradyear</th>\n",
       "      <th>gender</th>\n",
       "      <th>age</th>\n",
       "      <th>friends</th>\n",
       "      <th>basketball</th>\n",
       "      <th>football</th>\n",
       "      <th>soccer</th>\n",
       "      <th>softball</th>\n",
       "      <th>volleyball</th>\n",
       "      <th>swimming</th>\n",
       "      <th>...</th>\n",
       "      <th>blonde</th>\n",
       "      <th>mall</th>\n",
       "      <th>shopping</th>\n",
       "      <th>clothes</th>\n",
       "      <th>hollister</th>\n",
       "      <th>abercrombie</th>\n",
       "      <th>die</th>\n",
       "      <th>death</th>\n",
       "      <th>drunk</th>\n",
       "      <th>drugs</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2006</td>\n",
       "      <td>M</td>\n",
       "      <td>18.982</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2006</td>\n",
       "      <td>F</td>\n",
       "      <td>18.801</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2006</td>\n",
       "      <td>M</td>\n",
       "      <td>18.335</td>\n",
       "      <td>69</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2006</td>\n",
       "      <td>F</td>\n",
       "      <td>18.875</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2006</td>\n",
       "      <td>NaN</td>\n",
       "      <td>18.995</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 40 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   gradyear gender     age  friends  basketball  football  soccer  softball  \\\n",
       "0      2006      M  18.982        7           0         0       0         0   \n",
       "1      2006      F  18.801        0           0         1       0         0   \n",
       "2      2006      M  18.335       69           0         1       0         0   \n",
       "3      2006      F  18.875        0           0         0       0         0   \n",
       "4      2006    NaN  18.995       10           0         0       0         0   \n",
       "\n",
       "   volleyball  swimming  ...    blonde  mall  shopping  clothes  hollister  \\\n",
       "0           0         0  ...         0     0         0        0          0   \n",
       "1           0         0  ...         0     1         0        0          0   \n",
       "2           0         0  ...         0     0         0        0          0   \n",
       "3           0         0  ...         0     0         0        0          0   \n",
       "4           0         0  ...         0     0         2        0          0   \n",
       "\n",
       "   abercrombie  die  death  drunk  drugs  \n",
       "0            0    0      0      0      0  \n",
       "1            0    0      0      0      0  \n",
       "2            0    0      1      0      0  \n",
       "3            0    0      0      0      0  \n",
       "4            0    0      0      1      1  \n",
       "\n",
       "[5 rows x 40 columns]"
      ]
     },
     "execution_count": 139,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_sns = pd.read_csv('snsdata.csv', sep=',')\n",
    "df_sns.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>basketball</th>\n",
       "      <th>football</th>\n",
       "      <th>soccer</th>\n",
       "      <th>softball</th>\n",
       "      <th>volleyball</th>\n",
       "      <th>swimming</th>\n",
       "      <th>cheerleading</th>\n",
       "      <th>baseball</th>\n",
       "      <th>tennis</th>\n",
       "      <th>sports</th>\n",
       "      <th>...</th>\n",
       "      <th>blonde</th>\n",
       "      <th>mall</th>\n",
       "      <th>shopping</th>\n",
       "      <th>clothes</th>\n",
       "      <th>hollister</th>\n",
       "      <th>abercrombie</th>\n",
       "      <th>die</th>\n",
       "      <th>death</th>\n",
       "      <th>drunk</th>\n",
       "      <th>drugs</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 36 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   basketball  football  soccer  softball  volleyball  swimming  cheerleading  \\\n",
       "0           0         0       0         0           0         0             0   \n",
       "1           0         1       0         0           0         0             0   \n",
       "2           0         1       0         0           0         0             0   \n",
       "3           0         0       0         0           0         0             0   \n",
       "4           0         0       0         0           0         0             0   \n",
       "\n",
       "   baseball  tennis  sports  ...    blonde  mall  shopping  clothes  \\\n",
       "0         0       0       0  ...         0     0         0        0   \n",
       "1         0       0       0  ...         0     1         0        0   \n",
       "2         0       0       0  ...         0     0         0        0   \n",
       "3         0       0       0  ...         0     0         0        0   \n",
       "4         0       0       0  ...         0     0         2        0   \n",
       "\n",
       "   hollister  abercrombie  die  death  drunk  drugs  \n",
       "0          0            0    0      0      0      0  \n",
       "1          0            0    0      0      0      0  \n",
       "2          0            0    0      1      0      0  \n",
       "3          0            0    0      0      0      0  \n",
       "4          0            0    0      0      1      1  \n",
       "\n",
       "[5 rows x 36 columns]"
      ]
     },
     "execution_count": 140,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_sns = df_sns.drop(columns=['gradyear', 'age', 'friends', 'gender'])\n",
    "df_sns.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_sns = df_sns - df_sns.mean()\n",
    "df_sns = df_sns / df_sns.std()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Данные устроены так: \n",
    "* Год выпуска\n",
    "* Пол\n",
    "* Возраст\n",
    "* Количество друзей\n",
    "* 36 ключевых слов, которые встречаются в профилe facebook (интересы, сообщества, встречи)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Удалите все признаки кроме 36 ключевых слов.\n",
    "* Нормализуйте данные - из каждого столбца вычтите его среднее значение и поделите на стандартное отклонение.\n",
    "* Используйте метод k-means чтобы выделить 9 кластеров\n",
    "* Попробуйте проинтерпретировать каждый кластер проанализировав полученные центройды (Некоторые кластеры могут быть очень большие и очень маленькие - плохо интерпретируются)   \n",
    "КОММЕНТАРИИ НЕ СТИРАТЬ!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(30000, 36)"
      ]
     },
     "execution_count": 155,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X = df_sns.values\n",
    "X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 179,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 8\n"
     ]
    }
   ],
   "source": [
    "## MyKMeans\n",
    "model = MyKmeans(k=9, init='random')\n",
    "model.fit(X)\n",
    "df_sns_km = df_sns\n",
    "df_sns_km['label'] = model.labels\n",
    "print df_sns_km.label.min(), df_sns_km.label.max()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 183,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "label 0\n",
      "==========\n",
      "        basketball     football       soccer     softball   volleyball  \\\n",
      "count  3904.000000  3904.000000  3904.000000  3904.000000  3904.000000   \n",
      "mean      0.489668     0.468103     0.225175     0.262370     0.314292   \n",
      "std       1.755661     1.534665     1.549570     1.652281     1.738930   \n",
      "min      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "25%      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "50%      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "75%       0.910476     1.060031    -0.242870    -0.217924    -0.223666   \n",
      "max      29.492283    20.908147    29.193722    22.764155    21.653297   \n",
      "\n",
      "          swimming  cheerleading     baseball       tennis       sports  \\\n",
      "count  3904.000000   3904.000000  3904.000000  3904.000000  3904.000000   \n",
      "mean      0.331117      0.403248     0.262341     0.140248     0.479349   \n",
      "std       1.740572      1.806369     1.480216     1.317389     1.816400   \n",
      "min      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "25%      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "50%      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "75%      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "max      59.702484     17.291077    30.466308    21.109273    25.176241   \n",
      "\n",
      "        ...       shopping      clothes    hollister  abercrombie  \\\n",
      "count   ...    3904.000000  3904.000000  3904.000000  3904.000000   \n",
      "mean    ...       0.804763     1.005457     0.764677     0.784550   \n",
      "std     ...       1.642604     1.970740     2.230807     2.301982   \n",
      "min     ...      -0.487306    -0.314193    -0.201473    -0.183029   \n",
      "25%     ...      -0.487306    -0.314193    -0.201473    -0.183029   \n",
      "50%     ...       0.893164    -0.314193    -0.201473    -0.183029   \n",
      "75%     ...       0.893164     1.801583    -0.201473    -0.183029   \n",
      "max     ...      14.697870    16.612011    25.751622    28.433835   \n",
      "\n",
      "               die        death        drunk        drugs       labels   label  \n",
      "count  3904.000000  3904.000000  3904.000000  3904.000000  3904.000000  3904.0  \n",
      "mean      0.584991     0.428695     0.593370     0.751765     6.020748     0.0  \n",
      "std       1.952711     1.798309     1.981077     2.321750     0.202039     0.0  \n",
      "min      -0.294788    -0.261526    -0.220399    -0.174905     6.000000     0.0  \n",
      "25%      -0.294788    -0.261526    -0.220399    -0.174905     6.000000     0.0  \n",
      "50%      -0.294788    -0.261526    -0.220399    -0.174905     6.000000     0.0  \n",
      "75%       1.306452    -0.261526    -0.220399    -0.174905     6.000000     0.0  \n",
      "max      34.932494    31.790076    19.823463    46.131906     8.000000     0.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 1\n",
      "==========\n",
      "       basketball  football  soccer  softball  volleyball  swimming  \\\n",
      "count         0.0       0.0     0.0       0.0         0.0       0.0   \n",
      "mean          NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "std           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "min           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "25%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "50%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "75%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "max           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "\n",
      "       cheerleading  baseball  tennis  sports  ...    shopping  clothes  \\\n",
      "count           0.0       0.0     0.0     0.0  ...         0.0      0.0   \n",
      "mean            NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "std             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "min             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "25%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "50%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "75%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "max             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "\n",
      "       hollister  abercrombie  die  death  drunk  drugs  labels  label  \n",
      "count        0.0          0.0  0.0    0.0    0.0    0.0     0.0    0.0  \n",
      "mean         NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "std          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "min          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "25%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "50%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "75%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "max          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 2\n",
      "==========\n",
      "         basketball      football        soccer      softball    volleyball  \\\n",
      "count  25440.000000  25440.000000  25440.000000  25440.000000  25440.000000   \n",
      "mean      -0.078545     -0.072809     -0.034935     -0.040808     -0.048667   \n",
      "std        0.792305      0.868227      0.883259      0.847613      0.821106   \n",
      "min       -0.332212     -0.357691     -0.242870     -0.217924     -0.223666   \n",
      "25%       -0.332212     -0.357691     -0.242870     -0.217924     -0.223666   \n",
      "50%       -0.332212     -0.357691     -0.242870     -0.217924     -0.223666   \n",
      "75%       -0.332212     -0.357691     -0.242870     -0.217924     -0.223666   \n",
      "max       14.580035     10.984089     23.742501     16.004720     16.965376   \n",
      "\n",
      "           swimming  cheerleading      baseball        tennis        sports  \\\n",
      "count  25440.000000  25440.000000  25440.000000  25440.000000  25440.000000   \n",
      "mean      -0.052777     -0.061580     -0.041853     -0.024314     -0.073743   \n",
      "std        0.816518      0.787373      0.886217      0.924428      0.781892   \n",
      "min       -0.259966     -0.207324     -0.201127     -0.168936     -0.297118   \n",
      "25%       -0.259966     -0.207324     -0.201127     -0.168936     -0.297118   \n",
      "50%       -0.259966     -0.207324     -0.201127     -0.168936     -0.297118   \n",
      "75%       -0.259966     -0.207324     -0.201127     -0.168936     -0.297118   \n",
      "max       15.214214     17.291077     24.716164     24.978039     12.439561   \n",
      "\n",
      "        ...         shopping       clothes     hollister   abercrombie  \\\n",
      "count   ...     25440.000000  25440.000000  25440.000000  25440.000000   \n",
      "mean    ...        -0.124119     -0.155260     -0.114872     -0.118489   \n",
      "std     ...         0.790526      0.615869      0.547513      0.490707   \n",
      "min     ...        -0.487306     -0.314193     -0.201473     -0.183029   \n",
      "25%     ...        -0.487306     -0.314193     -0.201473     -0.183029   \n",
      "50%     ...        -0.487306     -0.314193     -0.201473     -0.183029   \n",
      "75%     ...        -0.487306     -0.314193     -0.201473     -0.183029   \n",
      "max     ...         6.415047      8.148909     11.333236      6.971187   \n",
      "\n",
      "                die         death         drunk         drugs        labels  \\\n",
      "count  25440.000000  25440.000000  25440.000000  25440.000000  25440.000000   \n",
      "mean      -0.096018     -0.074162     -0.093746     -0.117795      7.989937   \n",
      "std        0.696271      0.758673      0.685267      0.454041      0.141511   \n",
      "min       -0.294788     -0.261526     -0.220399     -0.174905      6.000000   \n",
      "25%       -0.294788     -0.261526     -0.220399     -0.174905      8.000000   \n",
      "50%       -0.294788     -0.261526     -0.220399     -0.174905      8.000000   \n",
      "75%       -0.294788     -0.261526     -0.220399     -0.174905      8.000000   \n",
      "max       17.318853     20.343075     12.307015     11.401798      8.000000   \n",
      "\n",
      "         label  \n",
      "count  25440.0  \n",
      "mean       2.0  \n",
      "std        0.0  \n",
      "min        2.0  \n",
      "25%        2.0  \n",
      "50%        2.0  \n",
      "75%        2.0  \n",
      "max        2.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 3\n",
      "==========\n",
      "       basketball  football   soccer  softball  volleyball  swimming  \\\n",
      "count    1.000000  1.000000  1.00000  1.000000    1.000000  1.000000   \n",
      "mean    -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "std           NaN       NaN      NaN       NaN         NaN       NaN   \n",
      "min     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "25%     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "50%     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "75%     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "max     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "\n",
      "       cheerleading  baseball    tennis    sports  ...    shopping   clothes  \\\n",
      "count      1.000000  1.000000  1.000000  1.000000  ...    1.000000  1.000000   \n",
      "mean      -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "std             NaN       NaN       NaN       NaN  ...         NaN       NaN   \n",
      "min       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "25%       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "50%       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "75%       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "max       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "\n",
      "       hollister  abercrombie       die      death      drunk     drugs  \\\n",
      "count   1.000000     1.000000  1.000000   1.000000   1.000000  1.000000   \n",
      "mean   -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "std          NaN          NaN       NaN        NaN        NaN       NaN   \n",
      "min    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "25%    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "50%    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "75%    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "max    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "\n",
      "       labels  label  \n",
      "count     1.0    1.0  \n",
      "mean      2.0    3.0  \n",
      "std       NaN    NaN  \n",
      "min       2.0    3.0  \n",
      "25%       2.0    3.0  \n",
      "50%       2.0    3.0  \n",
      "75%       2.0    3.0  \n",
      "max       2.0    3.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 4\n",
      "==========\n",
      "       basketball  football  soccer  softball  volleyball  swimming  \\\n",
      "count         0.0       0.0     0.0       0.0         0.0       0.0   \n",
      "mean          NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "std           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "min           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "25%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "50%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "75%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "max           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "\n",
      "       cheerleading  baseball  tennis  sports  ...    shopping  clothes  \\\n",
      "count           0.0       0.0     0.0     0.0  ...         0.0      0.0   \n",
      "mean            NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "std             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "min             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "25%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "50%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "75%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "max             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "\n",
      "       hollister  abercrombie  die  death  drunk  drugs  labels  label  \n",
      "count        0.0          0.0  0.0    0.0    0.0    0.0     0.0    0.0  \n",
      "mean         NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "std          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "min          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "25%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "50%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "75%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "max          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 5\n",
      "==========\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       basketball  football  soccer  softball  volleyball  swimming  \\\n",
      "count         0.0       0.0     0.0       0.0         0.0       0.0   \n",
      "mean          NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "std           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "min           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "25%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "50%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "75%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "max           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "\n",
      "       cheerleading  baseball  tennis  sports  ...    shopping  clothes  \\\n",
      "count           0.0       0.0     0.0     0.0  ...         0.0      0.0   \n",
      "mean            NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "std             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "min             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "25%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "50%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "75%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "max             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "\n",
      "       hollister  abercrombie  die  death  drunk  drugs  labels  label  \n",
      "count        0.0          0.0  0.0    0.0    0.0    0.0     0.0    0.0  \n",
      "mean         NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "std          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "min          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "25%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "50%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "75%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "max          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 6\n",
      "==========\n",
      "       basketball  football  soccer  softball  volleyball  swimming  \\\n",
      "count         0.0       0.0     0.0       0.0         0.0       0.0   \n",
      "mean          NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "std           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "min           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "25%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "50%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "75%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "max           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "\n",
      "       cheerleading  baseball  tennis  sports  ...    shopping  clothes  \\\n",
      "count           0.0       0.0     0.0     0.0  ...         0.0      0.0   \n",
      "mean            NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "std             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "min             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "25%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "50%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "75%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "max             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "\n",
      "       hollister  abercrombie  die  death  drunk  drugs  labels  label  \n",
      "count        0.0          0.0  0.0    0.0    0.0    0.0     0.0    0.0  \n",
      "mean         NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "std          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "min          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "25%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "50%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "75%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "max          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 7\n",
      "==========\n",
      "       basketball  football  soccer  softball  volleyball  swimming  \\\n",
      "count         0.0       0.0     0.0       0.0         0.0       0.0   \n",
      "mean          NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "std           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "min           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "25%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "50%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "75%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "max           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "\n",
      "       cheerleading  baseball  tennis  sports  ...    shopping  clothes  \\\n",
      "count           0.0       0.0     0.0     0.0  ...         0.0      0.0   \n",
      "mean            NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "std             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "min             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "25%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "50%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "75%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "max             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "\n",
      "       hollister  abercrombie  die  death  drunk  drugs  labels  label  \n",
      "count        0.0          0.0  0.0    0.0    0.0    0.0     0.0    0.0  \n",
      "mean         NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "std          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "min          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "25%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "50%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "75%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "max          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 8\n",
      "==========\n",
      "       basketball    football      soccer    softball  volleyball    swimming  \\\n",
      "count  655.000000  655.000000  655.000000  655.000000  655.000000  655.000000   \n",
      "mean     0.132610    0.034076    0.015127    0.021494    0.017291    0.073733   \n",
      "std      1.169625    0.983889    0.918637    1.074896    0.957261    1.041213   \n",
      "min     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "25%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "50%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "75%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "max      8.366599    9.566367    7.388839   13.300946   10.714816    9.411397   \n",
      "\n",
      "       cheerleading    baseball      tennis      sports  ...      shopping  \\\n",
      "count    655.000000  655.000000  655.000000  655.000000  ...    655.000000   \n",
      "mean      -0.011413    0.062238    0.108670    0.007525  ...      0.024838   \n",
      "std        1.078804    1.327333    1.455973    0.903636  ...      0.996596   \n",
      "min       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "25%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "50%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "75%       -0.207324   -0.201127   -0.168936   -0.297118  ...      0.893164   \n",
      "max       15.346810   20.882735   28.846804    6.071221  ...      7.795517   \n",
      "\n",
      "          clothes   hollister  abercrombie         die       death  \\\n",
      "count  655.000000  655.000000   655.000000  655.000000  655.000000   \n",
      "mean     0.034668   -0.095811    -0.073804    0.228365    0.304708   \n",
      "std      0.979587    0.705278     0.808252    1.292452    1.581115   \n",
      "min     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "25%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "50%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "75%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "max      6.033134    8.449559    14.125403   12.515132   20.343075   \n",
      "\n",
      "            drunk       drugs      labels  label  \n",
      "count  655.000000  655.000000  655.000000  655.0  \n",
      "mean     0.081789    0.094629    7.221374    8.0  \n",
      "std      1.181658    1.326177    0.975934    0.0  \n",
      "min     -0.220399   -0.174905    6.000000    8.0  \n",
      "25%     -0.220399   -0.174905    6.000000    8.0  \n",
      "50%     -0.220399   -0.174905    8.000000    8.0  \n",
      "75%     -0.220399   -0.174905    8.000000    8.0  \n",
      "max     14.812498   22.978501    8.000000    8.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n"
     ]
    }
   ],
   "source": [
    "for i in range(9):\n",
    "    print 'label', i\n",
    "    print '=' * 10\n",
    "    print df_sns_km[df_sns_km.label == i].describe()\n",
    "    print '=' * 10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 185,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 8\n"
     ]
    }
   ],
   "source": [
    "## MyKMeans++\n",
    "model = MyKmeans(k=9, init='k-means')\n",
    "model.fit(X)\n",
    "df_sns_km = df_sns\n",
    "df_sns_km['label'] = model.labels\n",
    "print df_sns_km.label.min(), df_sns_km.label.max()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 186,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "label 0\n",
      "==========\n",
      "       basketball  football   soccer  softball  volleyball  swimming  \\\n",
      "count    1.000000  1.000000  1.00000  1.000000    1.000000  1.000000   \n",
      "mean    -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "std           NaN       NaN      NaN       NaN         NaN       NaN   \n",
      "min     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "25%     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "50%     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "75%     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "max     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "\n",
      "       cheerleading  baseball    tennis    sports  ...    shopping   clothes  \\\n",
      "count      1.000000  1.000000  1.000000  1.000000  ...    1.000000  1.000000   \n",
      "mean      -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "std             NaN       NaN       NaN       NaN  ...         NaN       NaN   \n",
      "min       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "25%       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "50%       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "75%       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "max       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "\n",
      "       hollister  abercrombie       die      death      drunk     drugs  \\\n",
      "count   1.000000     1.000000  1.000000   1.000000   1.000000  1.000000   \n",
      "mean   -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "std          NaN          NaN       NaN        NaN        NaN       NaN   \n",
      "min    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "25%    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "50%    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "75%    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "max    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "\n",
      "       labels  label  \n",
      "count     1.0    1.0  \n",
      "mean      2.0    0.0  \n",
      "std       NaN    NaN  \n",
      "min       2.0    0.0  \n",
      "25%       2.0    0.0  \n",
      "50%       2.0    0.0  \n",
      "75%       2.0    0.0  \n",
      "max       2.0    0.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 1\n",
      "==========\n",
      "       basketball    football      soccer    softball  volleyball    swimming  \\\n",
      "count  471.000000  471.000000  471.000000  471.000000  471.000000  471.000000   \n",
      "mean     0.071464    0.051672   -0.055376   -0.017007   -0.044510    0.056252   \n",
      "std      1.058963    0.975960    0.672955    0.854091    0.741432    1.101484   \n",
      "min     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "25%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "50%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "75%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "max      8.366599    9.566367    4.118107    7.893398    6.026895    9.411397   \n",
      "\n",
      "       cheerleading    baseball      tennis      sports  ...      shopping  \\\n",
      "count    471.000000  471.000000  471.000000  471.000000  ...    471.000000   \n",
      "mean      -0.009182   -0.042418    0.052840   -0.040221  ...      0.054917   \n",
      "std        1.152368    0.636137    0.891411    0.819825  ...      1.023459   \n",
      "min       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "25%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "50%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "75%       -0.207324   -0.201127   -0.168936   -0.297118  ...      0.893164   \n",
      "max       15.346810    5.549017    9.502977    6.071221  ...      7.795517   \n",
      "\n",
      "          clothes   hollister  abercrombie         die       death  \\\n",
      "count  471.000000  471.000000   471.000000  471.000000  471.000000   \n",
      "mean     0.031698   -0.097391    -0.107082    0.286554    0.409254   \n",
      "std      1.006736    0.683263     0.566497    1.395376    1.759368   \n",
      "min     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "25%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "50%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "75%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "max      6.033134    8.449559     6.971187   12.515132   20.343075   \n",
      "\n",
      "            drunk       drugs      labels  label  \n",
      "count  471.000000  471.000000  471.000000  471.0  \n",
      "mean     0.066854    0.052451    7.150743    1.0  \n",
      "std      1.107350    0.944857    0.989624    0.0  \n",
      "min     -0.220399   -0.174905    6.000000    1.0  \n",
      "25%     -0.220399   -0.174905    6.000000    1.0  \n",
      "50%     -0.220399   -0.174905    8.000000    1.0  \n",
      "75%     -0.220399   -0.174905    8.000000    1.0  \n",
      "max      9.801532    8.507622    8.000000    1.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 2\n",
      "==========\n",
      "       basketball    football      soccer    softball  volleyball    swimming  \\\n",
      "count  684.000000  684.000000  684.000000  684.000000  684.000000  684.000000   \n",
      "mean     0.158323    0.172918    5.136624   -0.047950    0.052766    0.110487   \n",
      "std      1.079525    1.283948    2.752705    0.688567    1.149507    1.030643   \n",
      "min     -0.332212   -0.357691    3.027862   -0.217924   -0.223666   -0.259966   \n",
      "25%     -0.332212   -0.357691    3.027862   -0.217924   -0.223666   -0.259966   \n",
      "50%     -0.332212   -0.357691    4.118107   -0.217924   -0.223666   -0.259966   \n",
      "75%     -0.332212   -0.357691    6.298595   -0.217924   -0.223666   -0.259966   \n",
      "max      7.123912   12.401812   29.193722    7.893398   12.277456    9.411397   \n",
      "\n",
      "       cheerleading    baseball      tennis      sports  ...      shopping  \\\n",
      "count    684.000000  684.000000  684.000000  684.000000  ...    684.000000   \n",
      "mean      -0.019719    0.009039    0.096900    0.304957  ...      0.196874   \n",
      "std        0.934044    0.799304    1.186773    1.399767  ...      1.111763   \n",
      "min       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "25%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "50%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "75%       -0.207324   -0.201127   -0.168936   -0.297118  ...      0.893164   \n",
      "max       11.458277    7.465732   15.306125   10.316781  ...      6.415047   \n",
      "\n",
      "          clothes   hollister  abercrombie         die       death  \\\n",
      "count  684.000000  684.000000   684.000000  684.000000  684.000000   \n",
      "mean    -0.048174   -0.007541    -0.041827   -0.006846   -0.017189   \n",
      "std      0.764558    0.787268     0.697041    0.828574    0.818073   \n",
      "min     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "25%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "50%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "75%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "max      6.033134    5.565881     3.394079    4.508932    6.606674   \n",
      "\n",
      "            drunk       drugs      labels  label  \n",
      "count  684.000000  684.000000  684.000000  684.0  \n",
      "mean    -0.055565   -0.043736    7.602339    2.0  \n",
      "std      0.779081    0.622478    0.798824    0.0  \n",
      "min     -0.220399   -0.174905    6.000000    2.0  \n",
      "25%     -0.220399   -0.174905    8.000000    2.0  \n",
      "50%     -0.220399   -0.174905    8.000000    2.0  \n",
      "75%     -0.220399   -0.174905    8.000000    2.0  \n",
      "max      7.296049    5.613447    8.000000    2.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 3\n",
      "==========\n",
      "         basketball      football        soccer      softball    volleyball  \\\n",
      "count  19833.000000  19833.000000  19833.000000  19833.000000  19833.000000   \n",
      "mean      -0.184340     -0.185060     -0.152662     -0.141513     -0.131167   \n",
      "std        0.497392      0.568857      0.363717      0.392725      0.452708   \n",
      "min       -0.332212     -0.357691     -0.242870     -0.217924     -0.223666   \n",
      "25%       -0.332212     -0.357691     -0.242870     -0.217924     -0.223666   \n",
      "50%       -0.332212     -0.357691     -0.242870     -0.217924     -0.223666   \n",
      "75%       -0.332212     -0.357691     -0.242870     -0.217924     -0.223666   \n",
      "max        3.395850      3.895476      1.937618      3.837737      4.464255   \n",
      "\n",
      "           swimming  cheerleading      baseball        tennis        sports  \\\n",
      "count  19833.000000  19833.000000  19833.000000  19833.000000  19833.000000   \n",
      "mean      -0.102751     -0.115762     -0.138889     -0.039216     -0.143848   \n",
      "std        0.679715      0.538619      0.370518      0.875102      0.594718   \n",
      "min       -0.259966     -0.207324     -0.201127     -0.168936     -0.297118   \n",
      "25%       -0.259966     -0.207324     -0.201127     -0.168936     -0.297118   \n",
      "50%       -0.259966     -0.207324     -0.201127     -0.168936     -0.297118   \n",
      "75%       -0.259966     -0.207324     -0.201127     -0.168936     -0.297118   \n",
      "max       11.345669      9.514010      3.632302     24.978039      6.071221   \n",
      "\n",
      "        ...         shopping       clothes     hollister   abercrombie  \\\n",
      "count   ...     19833.000000  19833.000000  19833.000000  19833.000000   \n",
      "mean    ...        -0.253991     -0.203886     -0.159017     -0.150383   \n",
      "std     ...         0.580552      0.510837      0.347323      0.340171   \n",
      "min     ...        -0.487306     -0.314193     -0.201473     -0.183029   \n",
      "25%     ...        -0.487306     -0.314193     -0.201473     -0.183029   \n",
      "50%     ...        -0.487306     -0.314193     -0.201473     -0.183029   \n",
      "75%     ...        -0.487306     -0.314193     -0.201473     -0.183029   \n",
      "max     ...         3.654106      6.033134      2.682204      3.394079   \n",
      "\n",
      "                die         death         drunk         drugs        labels  \\\n",
      "count  19833.000000  19833.000000  19833.000000  19833.000000  19833.000000   \n",
      "mean      -0.097065     -0.080987     -0.092428     -0.114637      7.995865   \n",
      "std        0.711586      0.756369      0.696652      0.447625      0.090843   \n",
      "min       -0.294788     -0.261526     -0.220399     -0.174905      6.000000   \n",
      "25%       -0.294788     -0.261526     -0.220399     -0.174905      8.000000   \n",
      "50%       -0.294788     -0.261526     -0.220399     -0.174905      8.000000   \n",
      "75%       -0.294788     -0.261526     -0.220399     -0.174905      8.000000   \n",
      "max       15.717613     20.343075     12.307015      8.507622      8.000000   \n",
      "\n",
      "         label  \n",
      "count  19833.0  \n",
      "mean       3.0  \n",
      "std        0.0  \n",
      "min        3.0  \n",
      "25%        3.0  \n",
      "50%        3.0  \n",
      "75%        3.0  \n",
      "max        3.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 4\n",
      "==========\n",
      "       basketball    football      soccer    softball  volleyball    swimming  \\\n",
      "count  935.000000  935.000000  935.000000  935.000000  935.000000  935.000000   \n",
      "mean     0.345618    0.365575    0.035813    0.129084    0.098890    0.251013   \n",
      "std      1.581023    1.375858    0.934293    1.367999    1.102776    1.371758   \n",
      "min     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "25%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "50%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "75%      0.910476    1.060031   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "max     27.006908   15.237257   10.659571   22.764155   10.714816   19.082760   \n",
      "\n",
      "       cheerleading    baseball      tennis      sports  ...      shopping  \\\n",
      "count    935.000000  935.000000  935.000000  935.000000  ...    935.000000   \n",
      "mean       0.156577    0.221165    0.131048    0.772218  ...      0.240578   \n",
      "std        1.291697    1.343098    1.118931    2.083480  ...      1.267232   \n",
      "min       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "25%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "50%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "75%       -0.207324   -0.201127   -0.168936    1.825662  ...      0.893164   \n",
      "max       11.458277   22.799449   13.371743   25.176241  ...     10.556459   \n",
      "\n",
      "          clothes   hollister  abercrombie         die       death  \\\n",
      "count  935.000000  935.000000   935.000000  935.000000  935.000000   \n",
      "mean     1.201924    0.180961     0.276065    1.741441    0.913781   \n",
      "std      2.140429    1.472556     1.436381    3.226817    2.660763   \n",
      "min     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "25%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "50%     -0.314193   -0.201473    -0.183029    1.306452   -0.261526   \n",
      "75%      1.801583   -0.201473    -0.183029    2.907692    2.027874   \n",
      "max     12.380460   25.751622    10.548295   34.932494   31.790076   \n",
      "\n",
      "            drunk       drugs      labels  label  \n",
      "count  935.000000  935.000000  935.000000  935.0  \n",
      "mean     1.864377    2.926661    6.026738    4.0  \n",
      "std      3.090008    3.837641    0.227479    0.0  \n",
      "min     -0.220399   -0.174905    6.000000    4.0  \n",
      "25%     -0.220399   -0.174905    6.000000    4.0  \n",
      "50%      2.285084    2.719271    6.000000    4.0  \n",
      "75%      2.285084    5.613447    6.000000    4.0  \n",
      "max     19.823463   46.131906    8.000000    4.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 5\n",
      "==========\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       basketball    football      soccer    softball  volleyball    swimming  \\\n",
      "count  845.000000  845.000000  845.000000  845.000000  845.000000  845.000000   \n",
      "mean     0.116332    0.214431    0.024208    0.004457    0.172080    0.248210   \n",
      "std      1.038043    1.228161    0.825859    0.816845    1.310718    1.411875   \n",
      "min     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "25%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "50%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "75%     -0.332212    1.060031   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "max      7.123912   10.984089   10.659571    6.541511   16.965376   22.951305   \n",
      "\n",
      "       cheerleading    baseball      tennis      sports  ...      shopping  \\\n",
      "count    845.000000  845.000000  845.000000  845.000000  ...    845.000000   \n",
      "mean       0.411620    0.014361    0.133240    0.089755  ...      0.795143   \n",
      "std        1.681609    0.770310    1.457844    1.001696  ...      1.476679   \n",
      "min       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "25%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "50%       -0.207324   -0.201127   -0.168936   -0.297118  ...      0.893164   \n",
      "75%       -0.207324   -0.201127   -0.168936   -0.297118  ...      0.893164   \n",
      "max       11.458277    7.465732   21.109273    6.071221  ...     10.556459   \n",
      "\n",
      "          clothes   hollister  abercrombie         die       death  \\\n",
      "count  845.000000  845.000000   845.000000  845.000000  845.000000   \n",
      "mean     0.579691    4.170113     4.003669    0.050094    0.087980   \n",
      "std      1.831118    2.868320     3.176303    0.907005    1.013201   \n",
      "min     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "25%     -0.314193    2.682204     3.394079   -0.294788   -0.261526   \n",
      "50%     -0.314193    2.682204     3.394079   -0.294788   -0.261526   \n",
      "75%      1.801583    5.565881     3.394079   -0.294788   -0.261526   \n",
      "max     16.612011   22.867944    28.433835    6.110172    6.606674   \n",
      "\n",
      "            drunk       drugs      labels  label  \n",
      "count  845.000000  845.000000  845.000000  845.0  \n",
      "mean     0.058317    0.037449    6.589349    5.0  \n",
      "std      1.026041    0.876732    0.912333    0.0  \n",
      "min     -0.220399   -0.174905    6.000000    5.0  \n",
      "25%     -0.220399   -0.174905    6.000000    5.0  \n",
      "50%     -0.220399   -0.174905    6.000000    5.0  \n",
      "75%     -0.220399   -0.174905    8.000000    5.0  \n",
      "max      9.801532    8.507622    8.000000    5.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 6\n",
      "==========\n",
      "        basketball     football       soccer     softball   volleyball  \\\n",
      "count  2235.000000  2235.000000  2235.000000  2235.000000  2235.000000   \n",
      "mean      1.467600     1.248427     0.004935     1.374097     1.156491   \n",
      "std       2.320954     2.176561     0.642781     2.803957     2.684375   \n",
      "min      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "25%      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "50%       0.910476     1.060031    -0.242870    -0.217924    -0.223666   \n",
      "75%       2.153163     2.477754    -0.242870     2.485850     1.338974   \n",
      "max      29.492283    20.908147     4.118107    20.060381    21.653297   \n",
      "\n",
      "          swimming  cheerleading     baseball       tennis       sports  \\\n",
      "count  2235.000000   2235.000000  2235.000000  2235.000000  2235.000000   \n",
      "mean      0.084481      0.102367     1.328814     0.105426     1.016441   \n",
      "std       1.041047      1.186019     2.891468     1.270715     2.139443   \n",
      "min      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "25%      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "50%      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "75%      -0.259966     -0.207324     1.715587    -0.168936     1.825662   \n",
      "max      11.345669     17.291077    30.466308    21.109273    25.176241   \n",
      "\n",
      "        ...       shopping      clothes    hollister  abercrombie  \\\n",
      "count   ...    2235.000000  2235.000000  2235.000000  2235.000000   \n",
      "mean    ...       0.021028     0.001044    -0.105995    -0.103004   \n",
      "std     ...       0.947288     0.916715     0.544158     0.550354   \n",
      "min     ...      -0.487306    -0.314193    -0.201473    -0.183029   \n",
      "25%     ...      -0.487306    -0.314193    -0.201473    -0.183029   \n",
      "50%     ...      -0.487306    -0.314193    -0.201473    -0.183029   \n",
      "75%     ...       0.893164    -0.314193    -0.201473    -0.183029   \n",
      "max     ...       5.034576    10.264684     5.565881     6.971187   \n",
      "\n",
      "               die        death        drunk        drugs       labels   label  \n",
      "count  2235.000000  2235.000000  2235.000000  2235.000000  2235.000000  2235.0  \n",
      "mean     -0.071976    -0.033098    -0.070182    -0.099799     7.523937     6.0  \n",
      "std       0.672219     0.767009     0.726757     0.499318     0.851947     0.0  \n",
      "min      -0.294788    -0.261526    -0.220399    -0.174905     6.000000     6.0  \n",
      "25%      -0.294788    -0.261526    -0.220399    -0.174905     8.000000     6.0  \n",
      "50%      -0.294788    -0.261526    -0.220399    -0.174905     8.000000     6.0  \n",
      "75%      -0.294788    -0.261526    -0.220399    -0.174905     8.000000     6.0  \n",
      "max       6.110172     8.896075     7.296049     5.613447     8.000000     6.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 7\n",
      "==========\n",
      "       basketball    football      soccer    softball  volleyball    swimming  \\\n",
      "count  582.000000  582.000000  582.000000  582.000000  582.000000  582.000000   \n",
      "mean    -0.084528    0.068600   -0.132347   -0.053003   -0.075994    0.045795   \n",
      "std      0.748671    1.029516    0.552149    0.736281    0.952355    0.974456   \n",
      "min     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "25%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "50%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "75%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "max      7.123912    9.566367    7.388839    9.245285   12.277456    9.411397   \n",
      "\n",
      "       cheerleading    baseball      tennis      sports  ...      shopping  \\\n",
      "count    582.000000  582.000000  582.000000  582.000000  ...    582.000000   \n",
      "mean      -0.117126   -0.115501    0.043780   -0.092865  ...     -0.046125   \n",
      "std        0.580246    0.427027    1.545678    0.909325  ...      0.969492   \n",
      "min       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "25%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "50%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "75%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "max        7.569743    3.632302   28.846804   14.562341  ...      5.034576   \n",
      "\n",
      "          clothes   hollister  abercrombie         die       death  \\\n",
      "count  582.000000  582.000000   582.000000  582.000000  582.000000   \n",
      "mean    -0.016094   -0.166790    -0.140005    0.007852    0.033500   \n",
      "std      0.905623    0.314615     0.443123    0.874288    1.002189   \n",
      "min     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "25%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "50%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "75%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "max      6.033134    2.682204     6.971187    7.711412    8.896075   \n",
      "\n",
      "            drunk       drugs      labels  label  \n",
      "count  582.000000  582.000000  582.000000  582.0  \n",
      "mean    -0.091250   -0.065503    7.608247    7.0  \n",
      "std      0.644573    0.712054    0.794430    0.0  \n",
      "min     -0.220399   -0.174905    6.000000    7.0  \n",
      "25%     -0.220399   -0.174905    8.000000    7.0  \n",
      "50%     -0.220399   -0.174905    8.000000    7.0  \n",
      "75%     -0.220399   -0.174905    8.000000    7.0  \n",
      "max      4.790567    8.507622    8.000000    7.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 8\n",
      "==========\n",
      "        basketball     football       soccer     softball   volleyball  \\\n",
      "count  4414.000000  4414.000000  4414.000000  4414.000000  4414.000000   \n",
      "mean     -0.031253     0.038975    -0.101341    -0.071832    -0.043470   \n",
      "std       0.756660     0.878236     0.458538     0.568157     0.703453   \n",
      "min      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "25%      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "50%      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "75%      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "max       8.366599     6.730921     4.118107     6.541511     9.152175   \n",
      "\n",
      "          swimming  cheerleading     baseball       tennis       sports  \\\n",
      "count  4414.000000   4414.000000  4414.000000  4414.000000  4414.000000   \n",
      "mean      0.288677      0.375868    -0.079976     0.043171    -0.079743   \n",
      "std       1.676932      1.822338     0.536204     1.107641     0.728485   \n",
      "min      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "25%      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "50%      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "75%      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "max      59.702484     17.291077     7.465732    19.174891     8.194001   \n",
      "\n",
      "        ...       shopping      clothes    hollister  abercrombie  \\\n",
      "count   ...    4414.000000  4414.000000  4414.000000  4414.000000   \n",
      "mean    ...       0.897230     0.555797    -0.034881    -0.060658   \n",
      "std     ...       1.552440     1.554010     0.672865     0.650272   \n",
      "min     ...      -0.487306    -0.314193    -0.201473    -0.183029   \n",
      "25%     ...      -0.487306    -0.314193    -0.201473    -0.183029   \n",
      "50%     ...       0.893164    -0.314193    -0.201473    -0.183029   \n",
      "75%     ...       2.273635     1.801583    -0.201473    -0.183029   \n",
      "max     ...      14.697870    16.612011     2.682204     3.394079   \n",
      "\n",
      "               die        death        drunk        drugs       labels   label  \n",
      "count  4414.000000  4414.000000  4414.000000  4414.000000  4414.000000  4414.0  \n",
      "mean      0.061446     0.121770     0.054898    -0.051637     7.227458     8.0  \n",
      "std       0.893323     1.177447     1.007593     0.634335     0.973898     0.0  \n",
      "min      -0.294788    -0.261526    -0.220399    -0.174905     6.000000     8.0  \n",
      "25%      -0.294788    -0.261526    -0.220399    -0.174905     6.000000     8.0  \n",
      "50%      -0.294788    -0.261526    -0.220399    -0.174905     8.000000     8.0  \n",
      "75%      -0.294788    -0.261526    -0.220399    -0.174905     8.000000     8.0  \n",
      "max       9.312652    20.343075     9.801532     5.613447     8.000000     8.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n"
     ]
    }
   ],
   "source": [
    "for i in range(9):\n",
    "    print 'label', i\n",
    "    print '=' * 10\n",
    "    print df_sns_km[df_sns_km.label == i].describe()\n",
    "    print '=' * 10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 205,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 6\n"
     ]
    }
   ],
   "source": [
    "## MyKMeans с MiniBatchMyKMeans\n",
    "model = MiniBatchKMeans(k=9, init='random', batch_size=10000, max_iter=300)\n",
    "model.fit(X)\n",
    "df_sns_km = df_sns\n",
    "df_sns_km['label'] = model.labels\n",
    "print df_sns_km.label.min(), df_sns_km.label.max()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 206,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "label 0\n",
      "==========\n",
      "        basketball     football       soccer     softball   volleyball  \\\n",
      "count  4216.000000  4216.000000  4216.000000  4216.000000  4216.000000   \n",
      "mean      0.498703     0.452724     0.222604     0.251196     0.290790   \n",
      "std       1.759727     1.518002     1.535745     1.634211     1.680542   \n",
      "min      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "25%      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "50%      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "75%       0.910476     1.060031    -0.242870    -0.217924    -0.223666   \n",
      "max      29.492283    20.908147    29.193722    22.764155    21.653297   \n",
      "\n",
      "          swimming  cheerleading     baseball       tennis       sports  \\\n",
      "count  4216.000000   4216.000000  4216.000000  4216.000000  4216.000000   \n",
      "mean      0.306643      0.380199     0.255775     0.156367     0.453105   \n",
      "std       1.684108      1.769936     1.488911     1.427604     1.770894   \n",
      "min      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "25%      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "50%      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "75%      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "max      59.702484     17.291077    30.466308    28.846804    25.176241   \n",
      "\n",
      "        ...       shopping      clothes    hollister  abercrombie  \\\n",
      "count   ...    4216.000000  4216.000000  4216.000000  4216.000000   \n",
      "mean    ...       0.774633     0.953968     0.734217     0.755368   \n",
      "std     ...       1.621995     1.935236     2.193183     2.260925   \n",
      "min     ...      -0.487306    -0.314193    -0.201473    -0.183029   \n",
      "25%     ...      -0.487306    -0.314193    -0.201473    -0.183029   \n",
      "50%     ...       0.893164    -0.314193    -0.201473    -0.183029   \n",
      "75%     ...       0.893164     1.801583    -0.201473    -0.183029   \n",
      "max     ...      14.697870    16.612011    25.751622    28.433835   \n",
      "\n",
      "               die        death        drunk        drugs       labels   label  \n",
      "count  4216.000000  4216.000000  4216.000000  4216.000000  4216.000000  4216.0  \n",
      "mean      0.584830     0.452011     0.583661     0.733301     6.034867     0.0  \n",
      "std       1.933502     1.821603     1.965822     2.308566     0.261338     0.0  \n",
      "min      -0.294788    -0.261526    -0.220399    -0.174905     6.000000     0.0  \n",
      "25%      -0.294788    -0.261526    -0.220399    -0.174905     6.000000     0.0  \n",
      "50%      -0.294788    -0.261526    -0.220399    -0.174905     6.000000     0.0  \n",
      "75%       1.306452    -0.261526    -0.220399    -0.174905     6.000000     0.0  \n",
      "max      34.932494    31.790076    19.823463    46.131906     8.000000     0.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 1\n",
      "==========\n",
      "       basketball  football  soccer  softball  volleyball  swimming  \\\n",
      "count         0.0       0.0     0.0       0.0         0.0       0.0   \n",
      "mean          NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "std           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "min           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "25%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "50%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "75%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "max           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "\n",
      "       cheerleading  baseball  tennis  sports  ...    shopping  clothes  \\\n",
      "count           0.0       0.0     0.0     0.0  ...         0.0      0.0   \n",
      "mean            NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "std             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "min             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "25%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "50%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "75%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "max             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "\n",
      "       hollister  abercrombie  die  death  drunk  drugs  labels  label  \n",
      "count        0.0          0.0  0.0    0.0    0.0    0.0     0.0    0.0  \n",
      "mean         NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "std          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "min          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "25%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "50%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "75%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "max          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 2\n",
      "==========\n",
      "       basketball  football   soccer  softball  volleyball  swimming  \\\n",
      "count    1.000000  1.000000  1.00000  1.000000    1.000000  1.000000   \n",
      "mean    -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "std           NaN       NaN      NaN       NaN         NaN       NaN   \n",
      "min     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "25%     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "50%     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "75%     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "max     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "\n",
      "       cheerleading  baseball    tennis    sports  ...    shopping   clothes  \\\n",
      "count      1.000000  1.000000  1.000000  1.000000  ...    1.000000  1.000000   \n",
      "mean      -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "std             NaN       NaN       NaN       NaN  ...         NaN       NaN   \n",
      "min       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "25%       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "50%       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "75%       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "max       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "\n",
      "       hollister  abercrombie       die      death      drunk     drugs  \\\n",
      "count   1.000000     1.000000  1.000000   1.000000   1.000000  1.000000   \n",
      "mean   -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "std          NaN          NaN       NaN        NaN        NaN       NaN   \n",
      "min    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "25%    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "50%    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "75%    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "max    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "\n",
      "       labels  label  \n",
      "count     1.0    1.0  \n",
      "mean      2.0    2.0  \n",
      "std       NaN    NaN  \n",
      "min       2.0    2.0  \n",
      "25%       2.0    2.0  \n",
      "50%       2.0    2.0  \n",
      "75%       2.0    2.0  \n",
      "max       2.0    2.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 3\n",
      "==========\n",
      "       basketball  football  soccer  softball  volleyball  swimming  \\\n",
      "count         0.0       0.0     0.0       0.0         0.0       0.0   \n",
      "mean          NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "std           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "min           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "25%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "50%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "75%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "max           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "\n",
      "       cheerleading  baseball  tennis  sports  ...    shopping  clothes  \\\n",
      "count           0.0       0.0     0.0     0.0  ...         0.0      0.0   \n",
      "mean            NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "std             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "min             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "25%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "50%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "75%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "max             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "\n",
      "       hollister  abercrombie  die  death  drunk  drugs  labels  label  \n",
      "count        0.0          0.0  0.0    0.0    0.0    0.0     0.0    0.0  \n",
      "mean         NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "std          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "min          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "25%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "50%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "75%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "max          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 4\n",
      "==========\n",
      "       basketball  football  soccer  softball  volleyball  swimming  \\\n",
      "count         0.0       0.0     0.0       0.0         0.0       0.0   \n",
      "mean          NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "std           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "min           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "25%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "50%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "75%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "max           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "\n",
      "       cheerleading  baseball  tennis  sports  ...    shopping  clothes  \\\n",
      "count           0.0       0.0     0.0     0.0  ...         0.0      0.0   \n",
      "mean            NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "std             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "min             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "25%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "50%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "75%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "max             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "\n",
      "       hollister  abercrombie  die  death  drunk  drugs  labels  label  \n",
      "count        0.0          0.0  0.0    0.0    0.0    0.0     0.0    0.0  \n",
      "mean         NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "std          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "min          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "25%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "50%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "75%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "max          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 5\n",
      "==========\n",
      "       basketball  football  soccer  softball  volleyball  swimming  \\\n",
      "count         0.0       0.0     0.0       0.0         0.0       0.0   \n",
      "mean          NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "std           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "min           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "25%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "50%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "75%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "max           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "\n",
      "       cheerleading  baseball  tennis  sports  ...    shopping  clothes  \\\n",
      "count           0.0       0.0     0.0     0.0  ...         0.0      0.0   \n",
      "mean            NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "std             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "min             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "25%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "50%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "75%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "max             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "\n",
      "       hollister  abercrombie  die  death  drunk  drugs  labels  label  \n",
      "count        0.0          0.0  0.0    0.0    0.0    0.0     0.0    0.0  \n",
      "mean         NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "std          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "min          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "25%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "50%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "75%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "max          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 6\n",
      "==========\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "         basketball      football        soccer      softball    volleyball  \\\n",
      "count  25783.000000  25783.000000  25783.000000  25783.000000  25783.000000   \n",
      "mean      -0.081534     -0.074125     -0.036391     -0.041067     -0.047541   \n",
      "std        0.781015      0.864632      0.876670      0.845543      0.828102   \n",
      "min       -0.332212     -0.357691     -0.242870     -0.217924     -0.223666   \n",
      "25%       -0.332212     -0.357691     -0.242870     -0.217924     -0.223666   \n",
      "50%       -0.332212     -0.357691     -0.242870     -0.217924     -0.223666   \n",
      "75%       -0.332212     -0.357691     -0.242870     -0.217924     -0.223666   \n",
      "max       10.851974     10.984089     23.742501     16.004720     16.965376   \n",
      "\n",
      "           swimming  cheerleading      baseball        tennis        sports  \\\n",
      "count  25783.000000  25783.000000  25783.000000  25783.000000  25783.000000   \n",
      "mean      -0.050207     -0.062162     -0.041816     -0.025562     -0.074080   \n",
      "std        0.825757      0.789881      0.888082      0.908690      0.782174   \n",
      "min       -0.259966     -0.207324     -0.201127     -0.168936     -0.297118   \n",
      "25%       -0.259966     -0.207324     -0.201127     -0.168936     -0.297118   \n",
      "50%       -0.259966     -0.207324     -0.201127     -0.168936     -0.297118   \n",
      "75%       -0.259966     -0.207324     -0.201127     -0.168936     -0.297118   \n",
      "max       15.214214     17.291077     24.716164     24.978039     12.439561   \n",
      "\n",
      "        ...         shopping       clothes     hollister   abercrombie  \\\n",
      "count   ...     25783.000000  25783.000000  25783.000000  25783.000000   \n",
      "mean    ...        -0.126648     -0.156061     -0.120050     -0.123510   \n",
      "std     ...         0.786943      0.614806      0.524042      0.468292   \n",
      "min     ...        -0.487306     -0.314193     -0.201473     -0.183029   \n",
      "25%     ...        -0.487306     -0.314193     -0.201473     -0.183029   \n",
      "50%     ...        -0.487306     -0.314193     -0.201473     -0.183029   \n",
      "75%     ...        -0.487306     -0.314193     -0.201473     -0.183029   \n",
      "max     ...         6.415047      8.148909      8.449559      6.971187   \n",
      "\n",
      "                die         death         drunk         drugs        labels  \\\n",
      "count  25783.000000  25783.000000  25783.000000  25783.000000  25783.000000   \n",
      "mean      -0.095992     -0.074435     -0.096014     -0.119902      7.991933   \n",
      "std        0.695610      0.758343      0.677010      0.435841      0.126768   \n",
      "min       -0.294788     -0.261526     -0.220399     -0.174905      6.000000   \n",
      "25%       -0.294788     -0.261526     -0.220399     -0.174905      8.000000   \n",
      "50%       -0.294788     -0.261526     -0.220399     -0.174905      8.000000   \n",
      "75%       -0.294788     -0.261526     -0.220399     -0.174905      8.000000   \n",
      "max       17.318853     20.343075     12.307015     11.401798      8.000000   \n",
      "\n",
      "         label  \n",
      "count  25783.0  \n",
      "mean       6.0  \n",
      "std        0.0  \n",
      "min        6.0  \n",
      "25%        6.0  \n",
      "50%        6.0  \n",
      "75%        6.0  \n",
      "max        6.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 7\n",
      "==========\n",
      "       basketball  football  soccer  softball  volleyball  swimming  \\\n",
      "count         0.0       0.0     0.0       0.0         0.0       0.0   \n",
      "mean          NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "std           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "min           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "25%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "50%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "75%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "max           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "\n",
      "       cheerleading  baseball  tennis  sports  ...    shopping  clothes  \\\n",
      "count           0.0       0.0     0.0     0.0  ...         0.0      0.0   \n",
      "mean            NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "std             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "min             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "25%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "50%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "75%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "max             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "\n",
      "       hollister  abercrombie  die  death  drunk  drugs  labels  label  \n",
      "count        0.0          0.0  0.0    0.0    0.0    0.0     0.0    0.0  \n",
      "mean         NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "std          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "min          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "25%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "50%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "75%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "max          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 8\n",
      "==========\n",
      "       basketball  football  soccer  softball  volleyball  swimming  \\\n",
      "count         0.0       0.0     0.0       0.0         0.0       0.0   \n",
      "mean          NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "std           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "min           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "25%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "50%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "75%           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "max           NaN       NaN     NaN       NaN         NaN       NaN   \n",
      "\n",
      "       cheerleading  baseball  tennis  sports  ...    shopping  clothes  \\\n",
      "count           0.0       0.0     0.0     0.0  ...         0.0      0.0   \n",
      "mean            NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "std             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "min             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "25%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "50%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "75%             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "max             NaN       NaN     NaN     NaN  ...         NaN      NaN   \n",
      "\n",
      "       hollister  abercrombie  die  death  drunk  drugs  labels  label  \n",
      "count        0.0          0.0  0.0    0.0    0.0    0.0     0.0    0.0  \n",
      "mean         NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "std          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "min          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "25%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "50%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "75%          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "max          NaN          NaN  NaN    NaN    NaN    NaN     NaN    NaN  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n"
     ]
    }
   ],
   "source": [
    "for i in range(9):\n",
    "    print 'label', i\n",
    "    print '=' * 10\n",
    "    print df_sns_km[df_sns_km.label == i].describe()\n",
    "    print '=' * 10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 207,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 8\n"
     ]
    }
   ],
   "source": [
    "## MyKMeans++ с MiniBatchMyKMeans\n",
    "model = MiniBatchKMeans(k=9, init='k-means', batch_size=10000, max_iter=300)\n",
    "model.fit(X)\n",
    "df_sns_km = df_sns\n",
    "df_sns_km['label'] = model.labels\n",
    "print df_sns_km.label.min(), df_sns_km.label.max()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 208,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "label 0\n",
      "==========\n",
      "       basketball  football   soccer  softball  volleyball  swimming  \\\n",
      "count    1.000000  1.000000  1.00000  1.000000    1.000000  1.000000   \n",
      "mean    -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "std           NaN       NaN      NaN       NaN         NaN       NaN   \n",
      "min     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "25%     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "50%     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "75%     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "max     -0.332212  2.477754 -0.24287 -0.217924   -0.223666  1.674306   \n",
      "\n",
      "       cheerleading  baseball    tennis    sports  ...    shopping   clothes  \\\n",
      "count      1.000000  1.000000  1.000000  1.000000  ...    1.000000  1.000000   \n",
      "mean      -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "std             NaN       NaN       NaN       NaN  ...         NaN       NaN   \n",
      "min       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "25%       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "50%       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "75%       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "max       -0.207324 -0.201127 -0.168936 -0.297118  ...   -0.487306  1.801583   \n",
      "\n",
      "       hollister  abercrombie       die      death      drunk     drugs  \\\n",
      "count   1.000000     1.000000  1.000000   1.000000   1.000000  1.000000   \n",
      "mean   -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "std          NaN          NaN       NaN        NaN        NaN       NaN   \n",
      "min    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "25%    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "50%    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "75%    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "max    -0.201473    -0.183029  9.312652  13.474875  14.812498 -0.174905   \n",
      "\n",
      "       labels  label  \n",
      "count     1.0    1.0  \n",
      "mean      2.0    0.0  \n",
      "std       NaN    NaN  \n",
      "min       2.0    0.0  \n",
      "25%       2.0    0.0  \n",
      "50%       2.0    0.0  \n",
      "75%       2.0    0.0  \n",
      "max       2.0    0.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 1\n",
      "==========\n",
      "       basketball    football      soccer    softball  volleyball    swimming  \\\n",
      "count  918.000000  918.000000  918.000000  918.000000  918.000000  918.000000   \n",
      "mean     0.325682    0.357347    0.033848    0.117839    0.092947    0.249940   \n",
      "std      1.538048    1.375731    0.931819    1.358045    1.102316    1.379181   \n",
      "min     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "25%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "50%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "75%      0.910476    1.060031   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "max     27.006908   15.237257   10.659571   22.764155   10.714816   19.082760   \n",
      "\n",
      "       cheerleading    baseball      tennis      sports  ...      shopping  \\\n",
      "count    918.000000  918.000000  918.000000  918.000000  ...    918.000000   \n",
      "mean       0.150608    0.183051    0.130282    0.775834  ...      0.229997   \n",
      "std        1.293523    1.094701    1.124768    2.100056  ...      1.261000   \n",
      "min       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "25%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "50%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "75%       -0.207324   -0.201127   -0.168936    1.825662  ...      0.893164   \n",
      "max       11.458277    9.382446   13.371743   25.176241  ...     10.556459   \n",
      "\n",
      "          clothes   hollister  abercrombie         die       death  \\\n",
      "count  918.000000  918.000000   918.000000  918.000000  918.000000   \n",
      "mean     1.195429    0.175478     0.276774    1.712867    0.948016   \n",
      "std      2.146303    1.476221     1.421609    3.162886    2.760029   \n",
      "min     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "25%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "50%     -0.314193   -0.201473    -0.183029    1.306452   -0.261526   \n",
      "75%      1.801583   -0.201473    -0.183029    2.907692    2.027874   \n",
      "max     12.380460   25.751622    10.548295   34.932494   31.790076   \n",
      "\n",
      "            drunk       drugs      labels  label  \n",
      "count  918.000000  918.000000  918.000000  918.0  \n",
      "mean     1.842940    2.958876    6.031590    1.0  \n",
      "std      3.084056    3.805852    0.247306    0.0  \n",
      "min     -0.220399   -0.174905    6.000000    1.0  \n",
      "25%     -0.220399   -0.174905    6.000000    1.0  \n",
      "50%     -0.220399    2.719271    6.000000    1.0  \n",
      "75%      2.285084    5.613447    6.000000    1.0  \n",
      "max     19.823463   46.131906    8.000000    1.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 2\n",
      "==========\n",
      "       basketball    football      soccer    softball  volleyball    swimming  \\\n",
      "count  848.000000  848.000000  848.000000  848.000000  848.000000  848.000000   \n",
      "mean     0.129400    0.215751    0.027120    0.006859    0.170680    0.248693   \n",
      "std      1.100274    1.228762    0.825844    0.817274    1.306401    1.410440   \n",
      "min     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "25%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "50%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "75%     -0.332212    1.060031   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "max     10.851974   10.984089   10.659571    6.541511   16.965376   22.951305   \n",
      "\n",
      "       cheerleading    baseball      tennis      sports  ...      shopping  \\\n",
      "count    848.000000  848.000000  848.000000  848.000000  ...    848.000000   \n",
      "mean       0.404845    0.018120    0.136733    0.085883  ...      0.795490   \n",
      "std        1.670055    0.773401    1.457453    0.998487  ...      1.478645   \n",
      "min       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "25%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "50%       -0.207324   -0.201127   -0.168936   -0.297118  ...      0.893164   \n",
      "75%       -0.207324   -0.201127   -0.168936   -0.297118  ...      0.893164   \n",
      "max       11.458277    7.465732   21.109273    6.071221  ...     10.556459   \n",
      "\n",
      "          clothes   hollister  abercrombie         die       death  \\\n",
      "count  848.000000  848.000000   848.000000  848.000000  848.000000   \n",
      "mean     0.584014    4.161449     4.005731    0.056427    0.097543   \n",
      "std      1.832200    2.868071     3.177208    0.922624    1.044021   \n",
      "min     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "25%     -0.314193    2.682204     3.394079   -0.294788   -0.261526   \n",
      "50%     -0.314193    2.682204     3.394079   -0.294788   -0.261526   \n",
      "75%      1.801583    5.565881     3.394079   -0.294788   -0.261526   \n",
      "max     16.612011   22.867944    28.433835    6.110172    6.606674   \n",
      "\n",
      "            drunk       drugs      labels  label  \n",
      "count  848.000000  848.000000  848.000000  848.0  \n",
      "mean     0.057331    0.036698    6.587264    2.0  \n",
      "std      1.024356    0.875269    0.911388    0.0  \n",
      "min     -0.220399   -0.174905    6.000000    2.0  \n",
      "25%     -0.220399   -0.174905    6.000000    2.0  \n",
      "50%     -0.220399   -0.174905    6.000000    2.0  \n",
      "75%     -0.220399   -0.174905    8.000000    2.0  \n",
      "max      9.801532    8.507622    8.000000    2.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 3\n",
      "==========\n",
      "       basketball    football      soccer    softball  volleyball    swimming  \\\n",
      "count  302.000000  302.000000  302.000000  302.000000  302.000000  302.000000   \n",
      "mean     0.091619    0.050725   -0.058756   -0.007531   -0.021868    0.105112   \n",
      "std      1.181027    1.020429    0.666192    0.914326    0.805419    1.262551   \n",
      "min     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "25%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "50%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "75%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "max      8.366599    9.566367    4.118107    7.893398    6.026895    9.411397   \n",
      "\n",
      "       cheerleading    baseball      tennis      sports  ...      shopping  \\\n",
      "count    302.000000  302.000000  302.000000  302.000000  ...    302.000000   \n",
      "mean      -0.007747   -0.029765    0.068058   -0.022985  ...      0.010943   \n",
      "std        0.975795    0.649718    0.949137    0.847379  ...      0.947035   \n",
      "min       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "25%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "50%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "75%       -0.207324   -0.201127   -0.168936   -0.297118  ...      0.893164   \n",
      "max        7.569743    3.632302    9.502977    6.071221  ...      5.034576   \n",
      "\n",
      "          clothes   hollister  abercrombie         die       death  \\\n",
      "count  302.000000  302.000000   302.000000  302.000000  302.000000   \n",
      "mean     0.078137   -0.077341    -0.100116    0.320258    0.398003   \n",
      "std      1.032027    0.787370     0.612942    1.422177    1.469058   \n",
      "min     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "25%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "50%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "75%     -0.314193   -0.201473    -0.183029    0.906142   -0.261526   \n",
      "max      6.033134    8.449559     6.971187   12.515132    8.896075   \n",
      "\n",
      "            drunk       drugs      labels  label  \n",
      "count  302.000000  302.000000  302.000000  302.0  \n",
      "mean     0.053379    0.170096    7.099338    3.0  \n",
      "std      1.035296    1.648952    0.996705    0.0  \n",
      "min     -0.220399   -0.174905    6.000000    3.0  \n",
      "25%     -0.220399   -0.174905    6.000000    3.0  \n",
      "50%     -0.220399   -0.174905    8.000000    3.0  \n",
      "75%     -0.220399   -0.174905    8.000000    3.0  \n",
      "max      7.296049   22.978501    8.000000    3.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 4\n",
      "==========\n",
      "        basketball     football       soccer     softball   volleyball  \\\n",
      "count  2030.000000  2030.000000  2030.000000  2030.000000  2030.000000   \n",
      "mean      1.561203     1.190629     0.017070     1.456951     1.283551   \n",
      "std       2.373592     2.170049     0.662229     2.887291     2.792891   \n",
      "min      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "25%      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "50%       0.910476     1.060031    -0.242870    -0.217924    -0.223666   \n",
      "75%       2.153163     2.477754    -0.242870     2.485850     1.338974   \n",
      "max      29.492283    20.908147     4.118107    20.060381    21.653297   \n",
      "\n",
      "          swimming  cheerleading     baseball       tennis       sports  \\\n",
      "count  2030.000000   2030.000000  2030.000000  2030.000000  2030.000000   \n",
      "mean      0.101162      0.116402     1.417222     0.096922     1.125040   \n",
      "std       1.060785      1.208359     3.024106     1.222698     2.207842   \n",
      "min      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "25%      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "50%      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "75%      -0.259966     -0.207324     1.715587    -0.168936     1.825662   \n",
      "max      11.345669     17.291077    30.466308    21.109273    25.176241   \n",
      "\n",
      "        ...       shopping      clothes    hollister  abercrombie  \\\n",
      "count   ...    2030.000000  2030.000000  2030.000000  2030.000000   \n",
      "mean    ...       0.054682     0.033920    -0.106297    -0.100209   \n",
      "std     ...       0.977865     0.961523     0.546182     0.561039   \n",
      "min     ...      -0.487306    -0.314193    -0.201473    -0.183029   \n",
      "25%     ...      -0.487306    -0.314193    -0.201473    -0.183029   \n",
      "50%     ...      -0.487306    -0.314193    -0.201473    -0.183029   \n",
      "75%     ...       0.893164    -0.314193    -0.201473    -0.183029   \n",
      "max     ...       5.034576    10.264684     5.565881     6.971187   \n",
      "\n",
      "               die        death        drunk        drugs       labels   label  \n",
      "count  2030.000000  2030.000000  2030.000000  2030.000000  2030.000000  2030.0  \n",
      "mean     -0.062096    -0.011158    -0.074760    -0.087937     7.459113     4.0  \n",
      "std       0.687586     0.812765     0.701684     0.542030     0.888597     0.0  \n",
      "min      -0.294788    -0.261526    -0.220399    -0.174905     6.000000     4.0  \n",
      "25%      -0.294788    -0.261526    -0.220399    -0.174905     6.000000     4.0  \n",
      "50%      -0.294788    -0.261526    -0.220399    -0.174905     8.000000     4.0  \n",
      "75%      -0.294788    -0.261526    -0.220399    -0.174905     8.000000     4.0  \n",
      "max       6.110172     8.896075     7.296049     5.613447     8.000000     4.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 5\n",
      "==========\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "        basketball     football       soccer     softball   volleyball  \\\n",
      "count  4617.000000  4617.000000  4617.000000  4617.000000  4617.000000   \n",
      "mean     -0.036680     0.060839    -0.103077    -0.073571    -0.049024   \n",
      "std       0.741364     0.926788     0.454468     0.569219     0.666732   \n",
      "min      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "25%      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "50%      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "75%      -0.332212    -0.357691    -0.242870    -0.217924    -0.223666   \n",
      "max       8.366599     6.730921     4.118107     6.541511     7.589535   \n",
      "\n",
      "          swimming  cheerleading     baseball       tennis       sports  \\\n",
      "count  4617.000000   4617.000000  4617.000000  4617.000000  4617.000000   \n",
      "mean      0.285920      0.381389    -0.081981     0.092501    -0.075507   \n",
      "std       1.654013      1.829401     0.528661     1.355810     0.740486   \n",
      "min      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "25%      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "50%      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "75%      -0.259966     -0.207324    -0.201127    -0.168936    -0.297118   \n",
      "max      59.702484     17.291077     7.465732    24.978039     8.194001   \n",
      "\n",
      "        ...       shopping      clothes    hollister  abercrombie  \\\n",
      "count   ...    4617.000000  4617.000000  4617.000000  4617.000000   \n",
      "mean    ...       0.881504     0.555122    -0.035960    -0.064489   \n",
      "std     ...       1.542165     1.550691     0.670813     0.640364   \n",
      "min     ...      -0.487306    -0.314193    -0.201473    -0.183029   \n",
      "25%     ...      -0.487306    -0.314193    -0.201473    -0.183029   \n",
      "50%     ...       0.893164    -0.314193    -0.201473    -0.183029   \n",
      "75%     ...       2.273635     1.801583    -0.201473    -0.183029   \n",
      "max     ...      14.697870    16.612011     2.682204     3.394079   \n",
      "\n",
      "               die        death        drunk        drugs       labels   label  \n",
      "count  4617.000000  4617.000000  4617.000000  4617.000000  4617.000000  4617.0  \n",
      "mean      0.064164     0.121776     0.081323    -0.048281     7.231536     5.0  \n",
      "std       0.921563     1.174845     1.088375     0.642037     0.972932     0.0  \n",
      "min      -0.294788    -0.261526    -0.220399    -0.174905     6.000000     5.0  \n",
      "25%      -0.294788    -0.261526    -0.220399    -0.174905     6.000000     5.0  \n",
      "50%      -0.294788    -0.261526    -0.220399    -0.174905     8.000000     5.0  \n",
      "75%      -0.294788    -0.261526    -0.220399    -0.174905     8.000000     5.0  \n",
      "max      15.717613    20.343075     9.801532     5.613447     8.000000     5.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 6\n",
      "==========\n",
      "         basketball      football        soccer      softball    volleyball  \\\n",
      "count  20099.000000  20099.000000  20099.000000  20099.000000  20099.000000   \n",
      "mean      -0.174550     -0.168723     -0.152337     -0.133175     -0.129592   \n",
      "std        0.528308      0.623873      0.364170      0.431865      0.459412   \n",
      "min       -0.332212     -0.357691     -0.242870     -0.217924     -0.223666   \n",
      "25%       -0.332212     -0.357691     -0.242870     -0.217924     -0.223666   \n",
      "50%       -0.332212     -0.357691     -0.242870     -0.217924     -0.223666   \n",
      "75%       -0.332212     -0.357691     -0.242870     -0.217924     -0.223666   \n",
      "max        4.638537      5.313199      1.937618      5.189624      6.026895   \n",
      "\n",
      "           swimming  cheerleading      baseball        tennis        sports  \\\n",
      "count  20099.000000  20099.000000  20099.000000  20099.000000  20099.000000   \n",
      "mean      -0.105024     -0.119876     -0.130654     -0.048633     -0.141968   \n",
      "std        0.676036      0.521392      0.410493      0.797053      0.598879   \n",
      "min       -0.259966     -0.207324     -0.201127     -0.168936     -0.297118   \n",
      "25%       -0.259966     -0.207324     -0.201127     -0.168936     -0.297118   \n",
      "50%       -0.259966     -0.207324     -0.201127     -0.168936     -0.297118   \n",
      "75%       -0.259966     -0.207324     -0.201127     -0.168936     -0.297118   \n",
      "max       11.345669      7.569743      5.549017     21.109273      6.071221   \n",
      "\n",
      "        ...         shopping       clothes     hollister   abercrombie  \\\n",
      "count   ...     20099.000000  20099.000000  20099.000000  20099.000000   \n",
      "mean    ...        -0.258246     -0.209662     -0.159005     -0.150281   \n",
      "std     ...         0.573578      0.493617      0.347372      0.340696   \n",
      "min     ...        -0.487306     -0.314193     -0.201473     -0.183029   \n",
      "25%     ...        -0.487306     -0.314193     -0.201473     -0.183029   \n",
      "50%     ...        -0.487306     -0.314193     -0.201473     -0.183029   \n",
      "75%     ...        -0.487306     -0.314193     -0.201473     -0.183029   \n",
      "max     ...         3.654106      6.033134      2.682204      3.394079   \n",
      "\n",
      "                die         death         drunk         drugs        labels  \\\n",
      "count  20099.000000  20099.000000  20099.000000  20099.000000  20099.000000   \n",
      "mean      -0.094583     -0.081554     -0.095368     -0.116298      7.995920   \n",
      "std        0.729964      0.755005      0.683783      0.438213      0.090241   \n",
      "min       -0.294788     -0.261526     -0.220399     -0.174905      6.000000   \n",
      "25%       -0.294788     -0.261526     -0.220399     -0.174905      8.000000   \n",
      "50%       -0.294788     -0.261526     -0.220399     -0.174905      8.000000   \n",
      "75%       -0.294788     -0.261526     -0.220399     -0.174905      8.000000   \n",
      "max       17.318853     20.343075     12.307015      8.507622      8.000000   \n",
      "\n",
      "         label  \n",
      "count  20099.0  \n",
      "mean       6.0  \n",
      "std        0.0  \n",
      "min        6.0  \n",
      "25%        6.0  \n",
      "50%        6.0  \n",
      "75%        6.0  \n",
      "max        6.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 7\n",
      "==========\n",
      "       basketball    football      soccer    softball  volleyball    swimming  \\\n",
      "count  688.000000  688.000000  688.000000  688.000000  688.000000  688.000000   \n",
      "mean     0.160889    0.167773    5.129118   -0.046973    0.051159    0.111144   \n",
      "std      1.081847    1.269072    2.747932    0.688130    1.146349    1.034935   \n",
      "min     -0.332212   -0.357691    3.027862   -0.217924   -0.223666   -0.259966   \n",
      "25%     -0.332212   -0.357691    3.027862   -0.217924   -0.223666   -0.259966   \n",
      "50%     -0.332212   -0.357691    4.118107   -0.217924   -0.223666   -0.259966   \n",
      "75%     -0.332212   -0.357691    6.298595   -0.217924   -0.223666   -0.259966   \n",
      "max      7.123912   12.401812   29.193722    7.893398   12.277456    9.411397   \n",
      "\n",
      "       cheerleading    baseball      tennis      sports  ...      shopping  \\\n",
      "count    688.000000  688.000000  688.000000  688.000000  ...    688.000000   \n",
      "mean      -0.017984    0.013389    0.092543    0.286029  ...      0.196910   \n",
      "std        0.933809    0.808981    1.181809    1.343248  ...      1.112269   \n",
      "min       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "25%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "50%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "75%       -0.207324   -0.201127   -0.168936   -0.297118  ...      0.893164   \n",
      "max       11.458277    7.465732   15.306125   10.316781  ...      6.415047   \n",
      "\n",
      "          clothes   hollister  abercrombie         die       death  \\\n",
      "count  688.000000  688.000000   688.000000  688.000000  688.000000   \n",
      "mean    -0.055871   -0.008669    -0.042648   -0.006193   -0.021937   \n",
      "std      0.756137    0.785111     0.695092    0.827894    0.812202   \n",
      "min     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "25%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "50%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "75%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "max      6.033134    5.565881     3.394079    4.508932    6.606674   \n",
      "\n",
      "            drunk       drugs      labels  label  \n",
      "count  688.000000  688.000000  688.000000  688.0  \n",
      "mean    -0.045598   -0.048705    7.601744    7.0  \n",
      "std      0.803476    0.611728    0.799270    0.0  \n",
      "min     -0.220399   -0.174905    6.000000    7.0  \n",
      "25%     -0.220399   -0.174905    8.000000    7.0  \n",
      "50%     -0.220399   -0.174905    8.000000    7.0  \n",
      "75%     -0.220399   -0.174905    8.000000    7.0  \n",
      "max      7.296049    5.613447    8.000000    7.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n",
      "label 8\n",
      "==========\n",
      "       basketball    football      soccer    softball  volleyball    swimming  \\\n",
      "count  497.000000  497.000000  497.000000  497.000000  497.000000  497.000000   \n",
      "mean    -0.077173    0.098718   -0.124413   -0.041118   -0.066459    0.070845   \n",
      "std      0.779314    1.069923    0.583620    0.778540    0.994680    1.014574   \n",
      "min     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "25%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "50%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "75%     -0.332212   -0.357691   -0.242870   -0.217924   -0.223666   -0.259966   \n",
      "max      7.123912    9.566367    7.388839    9.245285   12.277456    9.411397   \n",
      "\n",
      "       cheerleading    baseball      tennis      sports  ...      shopping  \\\n",
      "count    497.000000  497.000000  497.000000  497.000000  ...    497.000000   \n",
      "mean      -0.109524   -0.112426    0.068483   -0.113457  ...     -0.029001   \n",
      "std        0.615713    0.438293    1.665091    0.900618  ...      0.996053   \n",
      "min       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "25%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "50%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "75%       -0.207324   -0.201127   -0.168936   -0.297118  ...     -0.487306   \n",
      "max        7.569743    3.632302   28.846804   14.562341  ...      5.034576   \n",
      "\n",
      "          clothes   hollister  abercrombie         die       death  \\\n",
      "count  497.000000  497.000000   497.000000  497.000000  497.000000   \n",
      "mean     0.005089   -0.166660    -0.139844    0.017727    0.056318   \n",
      "std      0.929275    0.315242     0.452232    0.915303    1.055464   \n",
      "min     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "25%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "50%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "75%     -0.314193   -0.201473    -0.183029   -0.294788   -0.261526   \n",
      "max      6.033134    2.682204     6.971187    7.711412    8.896075   \n",
      "\n",
      "            drunk       drugs      labels  label  \n",
      "count  497.000000  497.000000  497.000000  497.0  \n",
      "mean    -0.094369   -0.052616    7.569416    8.0  \n",
      "std      0.613534    0.759000    0.822877    0.0  \n",
      "min     -0.220399   -0.174905    6.000000    8.0  \n",
      "25%     -0.220399   -0.174905    8.000000    8.0  \n",
      "50%     -0.220399   -0.174905    8.000000    8.0  \n",
      "75%     -0.220399   -0.174905    8.000000    8.0  \n",
      "max      4.790567    8.507622    8.000000    8.0  \n",
      "\n",
      "[8 rows x 38 columns]\n",
      "==========\n"
     ]
    }
   ],
   "source": [
    "for i in range(9):\n",
    "    print 'label', i\n",
    "    print '=' * 10\n",
    "    print df_sns_km[df_sns_km.label == i].describe()\n",
    "    print '=' * 10"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Вывод\n",
    "\n",
    "И KMeans, и MiniBatchKMeans со случайной инициализацией центроидов плохо находят маленькие кластеры (или же игнорируют выбросы). В тоже время моя реализация оказалась чрезмерно чувствительной к крупным кластерам. Некоторые из 9 кластеров получались пустыми.\n",
    "\n",
    "В MiniBatchKMeans при маленьком batch_size находилось 2 крупных кластера, и больше ничего, пришлось подбирать более большой batch_size. (Причина понятна, при маленьком batch_size мал шанс того, что в батч на данной итерации попадёт достаточное число точек из маленького относительно других класа)\n",
    "\n",
    "На этих данных k-means++ в целом сработал лучше, самое адекватное разбиение получилось с применением MiniBatchKMeans с KMeans++. Выделились кластеры, похожие на те, которые мы получали в этих данных на лекции с использованием SKlearn KMeans"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.3"
  },
  "nav_menu": {},
  "toc": {
   "navigate_menu": true,
   "number_sections": false,
   "sideBar": true,
   "threshold": 6,
   "toc_cell": false,
   "toc_section_display": "block",
   "toc_window_display": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}