Handling_Imbalance_Dataset.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "86660b2a-b349-4199-a604-7120d7208cc0",
   "metadata": {},
   "source": [
    "Handling Imbalance Datset - suppose we have 1000 datapoints\n",
    "1000 = 900 (Yes)  and 100 (No)\n",
    "so the ratio is 900:100 or 9:10, this is an imbalanced datset, our model will get biased towards maximum number of datapoints so we need to fix them and make them equal.\n",
    "\n",
    "For this we use two techniques\n",
    "1.Upsampling - increasing the datapoints from minority\n",
    "2.Downsampling - decreasing the datapoints from majority"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "c5bfafd9-7f3d-4004-a796-da217cd37b3d",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "# Set the random seed for reproducibility\n",
    "np.random.seed(123)\n",
    "\n",
    "# Create a dataframe with two classes\n",
    "n_samples = 1000\n",
    "class_0_ratio = 0.9\n",
    "n_class_0 = int(n_samples * class_0_ratio)\n",
    "n_class_1 = n_samples - n_class_0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "4d982e58-06ee-488f-b5d7-4c48924d822a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(900, 100)"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "n_class_0,n_class_1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "f039fe30-2fae-4a71-9b16-3fe7c9321904",
   "metadata": {},
   "outputs": [],
   "source": [
    "## CREATE MY DATAFRAME WITH IMBALANCED DATASET\n",
    "class_0 = pd.DataFrame({\n",
    "    'feature_1': np.random.normal(loc=0, scale=1, size=n_class_0),\n",
    "    'feature_2': np.random.normal(loc=0, scale=1, size=n_class_0),\n",
    "    'target': [0] * n_class_0\n",
    "})\n",
    "\n",
    "class_1 = pd.DataFrame({\n",
    "    'feature_1': np.random.normal(loc=2, scale=1, size=n_class_1),\n",
    "    'feature_2': np.random.normal(loc=2, scale=1, size=n_class_1),\n",
    "    'target': [1] * n_class_1\n",
    "})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "1c5a0313-c825-4fb5-8ed8-291f0d1d3740",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.concat([class_0,class_1]).reset_index(drop = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "3195c19b-fea1-433a-9e3a-57de7a6a0d16",
   "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>feature_1</th>\n",
       "      <th>feature_2</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-1.085631</td>\n",
       "      <td>0.551302</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.997345</td>\n",
       "      <td>0.419589</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.282978</td>\n",
       "      <td>1.815652</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>-1.506295</td>\n",
       "      <td>-0.252750</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>-0.578600</td>\n",
       "      <td>-0.292004</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   feature_1  feature_2  target\n",
       "0  -1.085631   0.551302       0\n",
       "1   0.997345   0.419589       0\n",
       "2   0.282978   1.815652       0\n",
       "3  -1.506295  -0.252750       0\n",
       "4  -0.578600  -0.292004       0"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "3850bfe8-8bec-4bc0-b25f-6c30cae88e84",
   "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>feature_1</th>\n",
       "      <th>feature_2</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>995</th>\n",
       "      <td>1.376371</td>\n",
       "      <td>2.845701</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>996</th>\n",
       "      <td>2.239810</td>\n",
       "      <td>0.880077</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>997</th>\n",
       "      <td>1.131760</td>\n",
       "      <td>1.640703</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>998</th>\n",
       "      <td>2.902006</td>\n",
       "      <td>0.390305</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999</th>\n",
       "      <td>2.697490</td>\n",
       "      <td>2.013570</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     feature_1  feature_2  target\n",
       "995   1.376371   2.845701       1\n",
       "996   2.239810   0.880077       1\n",
       "997   1.131760   1.640703       1\n",
       "998   2.902006   0.390305       1\n",
       "999   2.697490   2.013570       1"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "3f543b2e-6168-4c50-9b01-aee490a8ed30",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    900\n",
       "1    100\n",
       "Name: target, dtype: int64"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['target'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "a93135c2-ebdd-4f05-ba2a-ae85888f2c5a",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_minority = df[df['target']==1]\n",
    "df_majority = df[df['target']==0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "e8f2b7a4-ccc6-47d9-810e-3537b82b9256",
   "metadata": {},
   "outputs": [],
   "source": [
    "#Perform upsampling\n",
    "from sklearn.utils import resample\n",
    "df_minority_upsampled = resample(df_minority,replace = True,\n",
    "                        n_samples = len(df_majority),\n",
    "                        random_state = 42\n",
    "                        )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "8023d017-ad6c-4597-bdde-3e7962df2104",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_upsampled = pd.concat([df_majority,df_minority_upsampled])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "c738b2ff-6a94-4335-8923-7badb156f7a6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    900\n",
       "1    900\n",
       "Name: target, dtype: int64"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_upsampled['target'].value_counts()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "893d8b4b-7822-4df2-952a-88d6b86933b6",
   "metadata": {},
   "outputs": [],
   "source": [
    "##Perform downsampling - But is is not good as we are losing datapoints"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "d6f6ae92-c353-41dc-a00f-32505b897159",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_minority = df[df['target']==1]\n",
    "df_majority = df[df['target']==0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "8fee8340-83ee-49aa-ae4e-0d8fc7a1260e",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.utils import resample\n",
    "df_majority_downsampled = resample(df_majority , replace = False,\n",
    "                          n_samples = len(df_minority),\n",
    "                          random_state = 42\n",
    "                          )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "42df8a50-e361-4c27-93fb-b5a6587b2ea5",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_downsampled = pd.concat([df_minority , df_majority_downsampled])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "fc5f3423-fe84-445b-8bde-28053be47a2f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1    100\n",
       "0    100\n",
       "Name: target, dtype: int64"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_downsampled['target'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8bc4382a-d9c6-499f-963e-66ae8dab6a67",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.10.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}