Suicidality_Prediction_TabNet.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Suicidality_TabNet.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "o2EFzvu-4kcC"
      },
      "source": [
        "# Using TabNet to predict Suicidality\n",
        "\n",
        "We want to predict suicidal ideation and attempts of the children through the deep learning model TabNet which is suit for tabular data.\n",
        "\n",
        "The TabNet training model works in four steps:\n",
        "1. Preprocessing the data\n",
        "2. Finding the hyperparameter which make the best auc performance (= hyperparameter-tuning)\n",
        "3. Training the TabNet model\n",
        "4. Extract important features\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5B2uqv2A5Zi5"
      },
      "source": [
        "# Get ready to work!\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BoQaCu_Y6wpb"
      },
      "source": [
        "## Import libraries and load dataset\n",
        "Import libraries we need and load the dataset we use"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aP04CfNL5qXG"
      },
      "source": [
        "1. import labraries"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "HeWRsJm0HlR0",
        "outputId": "14cfb28f-9686-4e14-db76-08224a9f2edb"
      },
      "source": [
        "!pip install pytorch_tabnet"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting pytorch_tabnet\n",
            "  Downloading https://files.pythonhosted.org/packages/94/e5/2a808d611a5d44e3c997c0d07362c04a56c70002208e00aec9eee3d923b5/pytorch_tabnet-3.1.1-py3-none-any.whl\n",
            "Requirement already satisfied: tqdm<5.0,>=4.36 in /usr/local/lib/python3.7/dist-packages (from pytorch_tabnet) (4.41.1)\n",
            "Requirement already satisfied: torch<2.0,>=1.2 in /usr/local/lib/python3.7/dist-packages (from pytorch_tabnet) (1.9.0+cu102)\n",
            "Requirement already satisfied: numpy<2.0,>=1.17 in /usr/local/lib/python3.7/dist-packages (from pytorch_tabnet) (1.19.5)\n",
            "Requirement already satisfied: scikit_learn>0.21 in /usr/local/lib/python3.7/dist-packages (from pytorch_tabnet) (0.22.2.post1)\n",
            "Requirement already satisfied: scipy>1.4 in /usr/local/lib/python3.7/dist-packages (from pytorch_tabnet) (1.4.1)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from torch<2.0,>=1.2->pytorch_tabnet) (3.7.4.3)\n",
            "Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.7/dist-packages (from scikit_learn>0.21->pytorch_tabnet) (1.0.1)\n",
            "Installing collected packages: pytorch-tabnet\n",
            "Successfully installed pytorch-tabnet-3.1.1\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KEpRWGWJ4anU"
      },
      "source": [
        "from torch import nn\n",
        "import torch\n",
        "from torch import tensor \n",
        "from torch.autograd import Variable\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt \n",
        "from pytorch_tabnet.tab_model import TabNetClassifier, TabNetRegressor\n",
        "from sklearn.metrics import roc_auc_score\n",
        "\"\"\"using stratified k fold instead of k fold\"\"\"\n",
        "from sklearn.model_selection import StratifiedKFold\n",
        "import torch\n",
        "import itertools\n",
        "from pytorch_tabnet.tab_model import TabNetClassifier\n",
        "from sklearn.metrics import roc_auc_score, accuracy_score\n",
        "from tqdm.notebook import tqdm\n",
        "from pathlib import Path\n",
        "import os\n",
        "from sklearn.metrics import confusion_matrix"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ogN21Bux5tgA"
      },
      "source": [
        "2. load data\n",
        "\n",
        "\n",
        "*   Load resting psychosocial, state fMRI, structural MRI, and diffusion MRI\n",
        "*   Create index arrays whose values are the positions(=index) of the features in column\n",
        "\n",
        "> The shape of the total data's subject is (5827, 5019).\n",
        "> \n",
        "> Train data has 5019 features of the 5510 subjects.\n",
        "> \n",
        "> Test data has 5019 features of the 317 subjects."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ge9PXsbiOL5t",
        "outputId": "d46248f3-cf0f-4728-b5d3-da47b2f54d32"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('content')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Mounted at content\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "53USOSv05m1m"
      },
      "source": [
        "# psychosocial + rsfMRI + strurctural MRI + diffusion MRI\n",
        "dataset_name = 'rsf(corr)_s_dMRI'\n",
        "\n",
        "#train_out = Path(os.getcwd()+'/sample_data/si_ppc_'+dataset_name+'_train.csv')\n",
        "#test_out =Path(os.getcwd()+'/sample_data/si_ppc_'+dataset_name+'_test.csv')\n",
        "train_out = '/content/sample_data/si_ppc_'+dataset_name+'_train.csv'\n",
        "test_out = '/content/sample_data/si_ppc_'+dataset_name+'_test.csv'\n",
        "\n",
        "# change form of the data file to csv\n",
        "train_data= pd.read_csv(train_out)\n",
        "test_data= pd.read_csv(test_out)\n",
        "\n",
        "# Set target (class) and not using features\n",
        "target ='Suicidalideation'\n",
        "unused_feat = ['abcd_site','kfold']"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Nzb2j9bg6gNP",
        "outputId": "bbec406f-6a62-4849-9ee5-9367b8bf76f0"
      },
      "source": [
        "# get the start column index of each feature\n",
        "start_psycho_index = np.where(test_data.columns.values == \"race.ethnicity_1\")[0][0]\n",
        "start_rsfmri_index = np.where(test_data.columns.values == \"rsfmri_cor_network.gordon_auditory_network.gordon_auditory\")[0][0]\n",
        "start_structmri_index = np.where(test_data.columns.values == \"lh_bankssts_area._.1\")[0][0]\n",
        "start_diffmri_index = np.where(test_data.columns.values == \"con_L.BSTS_L.CACG_count\")[0][0]\n",
        "\n",
        "psychosocial = list(test_data.columns[start_psycho_index:start_rsfmri_index])\n",
        "rsf_mri = list(test_data.columns[start_rsfmri_index:start_structmri_index])\n",
        "structural_mri = list(test_data.columns[start_structmri_index:start_diffmri_index])\n",
        "diffusion_mri = list(test_data.columns[start_diffmri_index:])\n",
        "\n",
        "Num_FOLDS  = 5\n",
        "# the number of feature that you want to show \n",
        "Num_feat = 20\n",
        "\n",
        "print(len(train_data), len(train_data.columns))\n",
        "print(len(test_data), len(test_data.columns))\n",
        "print(len(psychosocial))\n",
        "print(len(rsf_mri))\n",
        "print(len(structural_mri))\n",
        "print(len(diffusion_mri))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "2326 5019\n",
            "317 5019\n",
            "129\n",
            "416\n",
            "984\n",
            "3485\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9i-_bW4NQcra"
      },
      "source": [
        "## Define preprocessing function for Cross-validation\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6JlyRjwhQjLo"
      },
      "source": [
        "*   Divide all the subjects into sub-group by folds.\n",
        "\n",
        "*   In n-th fold, then we use n-th fold dataset as test(validation) data among k datasets.\n",
        "\n",
        "\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "53sO340CQmjr"
      },
      "source": [
        "def preprocessing (train_data, test_data, NUM_FOLDS):\n",
        "    test_data_processed= test_data.fillna(0).reset_index(drop=True)\n",
        "    train_data_processed = train_data.fillna(0).reset_index(drop=True)\n",
        "    \n",
        "    test_data_processed[\"kfold\"] = -1\n",
        "    train_data_processed[\"kfold\"] = -1\n",
        "\n",
        "    # reordering the data\n",
        "    train_data_processed = train_data_processed.sample(frac=1,random_state=2020).reset_index(drop=True)\n",
        "    \n",
        "    \"\"\"change k fold to stratified k fold\"\"\"\n",
        "    # kf = KFold(n_splits=NUM_FOLDS) \n",
        "    kf = StratifiedKFold(n_splits=NUM_FOLDS, shuffle=True, random_state =0)\n",
        "    \n",
        "    # Split the dataset into train(80%), validation(20%)\n",
        "    for fold, (trn_, val_) in enumerate(kf.split(X=train_data_processed, y=train_data_processed[target])):\n",
        "        # 'kfold' column : the fold order in cross validation\n",
        "        train_data_processed.loc[val_, 'kfold'] = fold\n",
        "    \n",
        "    print(\"done preprocessing\")\n",
        "    return train_data_processed, test_data_processed"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2EkUQZlPVifJ",
        "outputId": "a618b0f4-2db5-41b1-e3d7-908584dc442e"
      },
      "source": [
        "train_data_processed, test_data_processed = preprocessing (train_data, test_data, Num_FOLDS)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "done preprocessing\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 425
        },
        "id": "X3GkRim81dmc",
        "outputId": "5126ddfc-18e5-4132-e479-781beed336f7"
      },
      "source": [
        "train_data_processed.head(10)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>Unnamed: 0</th>\n",
              "      <th>X</th>\n",
              "      <th>subjectkey</th>\n",
              "      <th>abcd_site</th>\n",
              "      <th>Suicidalideation</th>\n",
              "      <th>race.ethnicity_1</th>\n",
              "      <th>race.ethnicity_2</th>\n",
              "      <th>race.ethnicity_3</th>\n",
              "      <th>race.ethnicity_4</th>\n",
              "      <th>race.ethnicity_5</th>\n",
              "      <th>sex_1</th>\n",
              "      <th>sex_2</th>\n",
              "      <th>high.educ_1</th>\n",
              "      <th>high.educ_2</th>\n",
              "      <th>high.educ_3</th>\n",
              "      <th>high.educ_4</th>\n",
              "      <th>high.educ_5</th>\n",
              "      <th>high.educ_6</th>\n",
              "      <th>high.educ_7</th>\n",
              "      <th>high.educ_8</th>\n",
              "      <th>high.educ_9</th>\n",
              "      <th>high.educ_10</th>\n",
              "      <th>high.educ_11</th>\n",
              "      <th>high.educ_12</th>\n",
              "      <th>high.educ_13</th>\n",
              "      <th>high.educ_14</th>\n",
              "      <th>high.educ_15</th>\n",
              "      <th>high.educ_16</th>\n",
              "      <th>high.educ_17</th>\n",
              "      <th>high.educ_18</th>\n",
              "      <th>high.educ_19</th>\n",
              "      <th>income_1</th>\n",
              "      <th>income_2</th>\n",
              "      <th>income_3</th>\n",
              "      <th>income_4</th>\n",
              "      <th>income_5</th>\n",
              "      <th>income_6</th>\n",
              "      <th>income_7</th>\n",
              "      <th>income_8</th>\n",
              "      <th>income_9</th>\n",
              "      <th>...</th>\n",
              "      <th>con_R.RMFG_R.TTG_count</th>\n",
              "      <th>con_R.RMFG_R.IN_count</th>\n",
              "      <th>con_R.RMFG_R.CER_count</th>\n",
              "      <th>con_R.SFG_R.SPG_count</th>\n",
              "      <th>con_R.SFG_R.STG_count</th>\n",
              "      <th>con_R.SFG_R.SMG_count</th>\n",
              "      <th>con_R.SFG_R.FP_count</th>\n",
              "      <th>con_R.SFG_R.TP_count</th>\n",
              "      <th>con_R.SFG_R.TTG_count</th>\n",
              "      <th>con_R.SFG_R.IN_count</th>\n",
              "      <th>con_R.SFG_R.CER_count</th>\n",
              "      <th>con_R.SPG_R.STG_count</th>\n",
              "      <th>con_R.SPG_R.SMG_count</th>\n",
              "      <th>con_R.SPG_R.FP_count</th>\n",
              "      <th>con_R.SPG_R.TP_count</th>\n",
              "      <th>con_R.SPG_R.TTG_count</th>\n",
              "      <th>con_R.SPG_R.IN_count</th>\n",
              "      <th>con_R.SPG_R.CER_count</th>\n",
              "      <th>con_R.STG_R.SMG_count</th>\n",
              "      <th>con_R.STG_R.FP_count</th>\n",
              "      <th>con_R.STG_R.TP_count</th>\n",
              "      <th>con_R.STG_R.TTG_count</th>\n",
              "      <th>con_R.STG_R.IN_count</th>\n",
              "      <th>con_R.STG_R.CER_count</th>\n",
              "      <th>con_R.SMG_R.FP_count</th>\n",
              "      <th>con_R.SMG_R.TP_count</th>\n",
              "      <th>con_R.SMG_R.TTG_count</th>\n",
              "      <th>con_R.SMG_R.IN_count</th>\n",
              "      <th>con_R.SMG_R.CER_count</th>\n",
              "      <th>con_R.FP_R.TP_count</th>\n",
              "      <th>con_R.FP_R.TTG_count</th>\n",
              "      <th>con_R.FP_R.IN_count</th>\n",
              "      <th>con_R.FP_R.CER_count</th>\n",
              "      <th>con_R.TP_R.TTG_count</th>\n",
              "      <th>con_R.TP_R.IN_count</th>\n",
              "      <th>con_R.TP_R.CER_count</th>\n",
              "      <th>con_R.TTG_R.IN_count</th>\n",
              "      <th>con_R.TTG_R.CER_count</th>\n",
              "      <th>con_R.IN_R.CER_count</th>\n",
              "      <th>kfold</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>1799</td>\n",
              "      <td>1799</td>\n",
              "      <td>NDARINVA1A1R65H</td>\n",
              "      <td>7</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>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>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>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>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.716253</td>\n",
              "      <td>-0.468683</td>\n",
              "      <td>-0.333610</td>\n",
              "      <td>0.996137</td>\n",
              "      <td>0.014790</td>\n",
              "      <td>0.465232</td>\n",
              "      <td>0.672203</td>\n",
              "      <td>0.103277</td>\n",
              "      <td>-0.536019</td>\n",
              "      <td>-0.592324</td>\n",
              "      <td>-0.100393</td>\n",
              "      <td>2.200748</td>\n",
              "      <td>-0.136701</td>\n",
              "      <td>1.607978</td>\n",
              "      <td>-0.368455</td>\n",
              "      <td>1.089732</td>\n",
              "      <td>0.462392</td>\n",
              "      <td>-0.927877</td>\n",
              "      <td>0.883990</td>\n",
              "      <td>0.331896</td>\n",
              "      <td>-0.769956</td>\n",
              "      <td>2.122436</td>\n",
              "      <td>-1.434421</td>\n",
              "      <td>-0.287827</td>\n",
              "      <td>-0.361203</td>\n",
              "      <td>0.889571</td>\n",
              "      <td>0.052236</td>\n",
              "      <td>-0.265021</td>\n",
              "      <td>-0.543032</td>\n",
              "      <td>-0.261251</td>\n",
              "      <td>-0.106171</td>\n",
              "      <td>0.968776</td>\n",
              "      <td>-0.13702</td>\n",
              "      <td>-0.189721</td>\n",
              "      <td>-0.503834</td>\n",
              "      <td>-0.274884</td>\n",
              "      <td>0.991228</td>\n",
              "      <td>-0.474512</td>\n",
              "      <td>0.572809</td>\n",
              "      <td>4</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>876</td>\n",
              "      <td>876</td>\n",
              "      <td>NDARINV4X6J651P</td>\n",
              "      <td>3</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",
              "      <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>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>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>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.041528</td>\n",
              "      <td>-0.658974</td>\n",
              "      <td>-0.139992</td>\n",
              "      <td>0.823889</td>\n",
              "      <td>0.100951</td>\n",
              "      <td>-0.192566</td>\n",
              "      <td>0.705836</td>\n",
              "      <td>1.767457</td>\n",
              "      <td>-0.768309</td>\n",
              "      <td>-1.114687</td>\n",
              "      <td>-0.670959</td>\n",
              "      <td>1.042577</td>\n",
              "      <td>0.402447</td>\n",
              "      <td>-0.323452</td>\n",
              "      <td>-0.131015</td>\n",
              "      <td>0.106278</td>\n",
              "      <td>-0.716357</td>\n",
              "      <td>0.018715</td>\n",
              "      <td>1.137880</td>\n",
              "      <td>-0.365029</td>\n",
              "      <td>0.324429</td>\n",
              "      <td>0.475110</td>\n",
              "      <td>-0.421332</td>\n",
              "      <td>-0.287827</td>\n",
              "      <td>-0.361203</td>\n",
              "      <td>-0.399832</td>\n",
              "      <td>0.677507</td>\n",
              "      <td>0.347327</td>\n",
              "      <td>0.216824</td>\n",
              "      <td>-0.261251</td>\n",
              "      <td>-0.106171</td>\n",
              "      <td>0.165055</td>\n",
              "      <td>-0.13702</td>\n",
              "      <td>-0.189721</td>\n",
              "      <td>0.536635</td>\n",
              "      <td>-0.274884</td>\n",
              "      <td>-0.990834</td>\n",
              "      <td>0.830664</td>\n",
              "      <td>-0.778556</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>783</td>\n",
              "      <td>783</td>\n",
              "      <td>NDARINV4BU3CBM5</td>\n",
              "      <td>21</td>\n",
              "      <td>1</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",
              "      <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>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",
              "      <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>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>...</td>\n",
              "      <td>0.183381</td>\n",
              "      <td>-0.986451</td>\n",
              "      <td>-0.527228</td>\n",
              "      <td>0.260770</td>\n",
              "      <td>1.651840</td>\n",
              "      <td>0.824031</td>\n",
              "      <td>0.255155</td>\n",
              "      <td>0.103277</td>\n",
              "      <td>1.554597</td>\n",
              "      <td>-0.011921</td>\n",
              "      <td>-0.528317</td>\n",
              "      <td>0.468051</td>\n",
              "      <td>-0.859858</td>\n",
              "      <td>-0.323452</td>\n",
              "      <td>-0.131015</td>\n",
              "      <td>-0.494721</td>\n",
              "      <td>-1.585831</td>\n",
              "      <td>-0.888436</td>\n",
              "      <td>0.134889</td>\n",
              "      <td>-0.365029</td>\n",
              "      <td>-0.378139</td>\n",
              "      <td>-0.195657</td>\n",
              "      <td>1.163577</td>\n",
              "      <td>0.176657</td>\n",
              "      <td>-0.361203</td>\n",
              "      <td>-0.399832</td>\n",
              "      <td>-0.225663</td>\n",
              "      <td>-1.203170</td>\n",
              "      <td>-0.868684</td>\n",
              "      <td>-0.261251</td>\n",
              "      <td>-0.106171</td>\n",
              "      <td>-0.192154</td>\n",
              "      <td>-0.13702</td>\n",
              "      <td>-0.189721</td>\n",
              "      <td>0.990738</td>\n",
              "      <td>-0.274884</td>\n",
              "      <td>-0.226914</td>\n",
              "      <td>0.830664</td>\n",
              "      <td>-0.665942</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>1425</td>\n",
              "      <td>1425</td>\n",
              "      <td>NDARINV83TT3MW4</td>\n",
              "      <td>16</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>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>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",
              "      <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>1</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.266436</td>\n",
              "      <td>-1.382521</td>\n",
              "      <td>-0.527228</td>\n",
              "      <td>2.334372</td>\n",
              "      <td>-0.674494</td>\n",
              "      <td>-0.162666</td>\n",
              "      <td>-0.108081</td>\n",
              "      <td>1.767457</td>\n",
              "      <td>-0.071438</td>\n",
              "      <td>0.088330</td>\n",
              "      <td>0.755456</td>\n",
              "      <td>-0.170310</td>\n",
              "      <td>0.277320</td>\n",
              "      <td>-0.323452</td>\n",
              "      <td>0.581304</td>\n",
              "      <td>1.435761</td>\n",
              "      <td>-0.407082</td>\n",
              "      <td>4.081169</td>\n",
              "      <td>-0.445791</td>\n",
              "      <td>-0.365029</td>\n",
              "      <td>-0.222763</td>\n",
              "      <td>-0.306136</td>\n",
              "      <td>-0.891322</td>\n",
              "      <td>0.176657</td>\n",
              "      <td>-0.361203</td>\n",
              "      <td>0.889571</td>\n",
              "      <td>1.747417</td>\n",
              "      <td>0.563219</td>\n",
              "      <td>3.690452</td>\n",
              "      <td>-0.261251</td>\n",
              "      <td>-0.106171</td>\n",
              "      <td>-0.549363</td>\n",
              "      <td>-0.13702</td>\n",
              "      <td>3.344864</td>\n",
              "      <td>1.630009</td>\n",
              "      <td>2.564233</td>\n",
              "      <td>-0.732753</td>\n",
              "      <td>-0.474512</td>\n",
              "      <td>1.924174</td>\n",
              "      <td>3</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>442</td>\n",
              "      <td>442</td>\n",
              "      <td>NDARINV2BW6G83G</td>\n",
              "      <td>8</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>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>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",
              "      <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>-0.041528</td>\n",
              "      <td>1.350143</td>\n",
              "      <td>0.440861</td>\n",
              "      <td>1.784503</td>\n",
              "      <td>-0.329852</td>\n",
              "      <td>-0.132766</td>\n",
              "      <td>-0.565488</td>\n",
              "      <td>-0.451449</td>\n",
              "      <td>-0.768309</td>\n",
              "      <td>0.805920</td>\n",
              "      <td>-0.480770</td>\n",
              "      <td>0.759874</td>\n",
              "      <td>1.177127</td>\n",
              "      <td>-0.323452</td>\n",
              "      <td>0.106424</td>\n",
              "      <td>-0.130479</td>\n",
              "      <td>1.728673</td>\n",
              "      <td>-0.888436</td>\n",
              "      <td>1.638119</td>\n",
              "      <td>-0.365029</td>\n",
              "      <td>0.999976</td>\n",
              "      <td>4.085416</td>\n",
              "      <td>0.722309</td>\n",
              "      <td>-0.752311</td>\n",
              "      <td>-0.361203</td>\n",
              "      <td>2.178974</td>\n",
              "      <td>1.302779</td>\n",
              "      <td>2.192223</td>\n",
              "      <td>-0.760133</td>\n",
              "      <td>-0.261251</td>\n",
              "      <td>-0.106171</td>\n",
              "      <td>0.432962</td>\n",
              "      <td>-0.13702</td>\n",
              "      <td>-0.189721</td>\n",
              "      <td>3.190712</td>\n",
              "      <td>-0.274884</td>\n",
              "      <td>-0.877278</td>\n",
              "      <td>-0.474512</td>\n",
              "      <td>4.739517</td>\n",
              "      <td>4</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5</th>\n",
              "      <td>1653</td>\n",
              "      <td>1653</td>\n",
              "      <td>NDARINV97EZB853</td>\n",
              "      <td>10</td>\n",
              "      <td>1</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>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>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>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>1</td>\n",
              "      <td>0</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.266436</td>\n",
              "      <td>0.653147</td>\n",
              "      <td>2.957891</td>\n",
              "      <td>0.982887</td>\n",
              "      <td>0.014790</td>\n",
              "      <td>1.392130</td>\n",
              "      <td>0.409866</td>\n",
              "      <td>0.103277</td>\n",
              "      <td>-0.536019</td>\n",
              "      <td>1.684439</td>\n",
              "      <td>1.468663</td>\n",
              "      <td>0.705157</td>\n",
              "      <td>0.356444</td>\n",
              "      <td>-0.323452</td>\n",
              "      <td>0.343864</td>\n",
              "      <td>0.215551</td>\n",
              "      <td>0.911718</td>\n",
              "      <td>-0.888436</td>\n",
              "      <td>1.969935</td>\n",
              "      <td>-0.365029</td>\n",
              "      <td>-0.418672</td>\n",
              "      <td>0.483002</td>\n",
              "      <td>-0.478775</td>\n",
              "      <td>-0.287827</td>\n",
              "      <td>-0.361203</td>\n",
              "      <td>0.889571</td>\n",
              "      <td>-0.434087</td>\n",
              "      <td>0.629949</td>\n",
              "      <td>-0.325930</td>\n",
              "      <td>-0.261251</td>\n",
              "      <td>-0.106171</td>\n",
              "      <td>-0.638665</td>\n",
              "      <td>-0.13702</td>\n",
              "      <td>-0.189721</td>\n",
              "      <td>0.646854</td>\n",
              "      <td>-0.274884</td>\n",
              "      <td>-0.319823</td>\n",
              "      <td>0.830664</td>\n",
              "      <td>3.162925</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>6</th>\n",
              "      <td>1316</td>\n",
              "      <td>1316</td>\n",
              "      <td>NDARINV7DY4B12E</td>\n",
              "      <td>11</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>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>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",
              "      <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>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.266436</td>\n",
              "      <td>-0.420004</td>\n",
              "      <td>0.440861</td>\n",
              "      <td>-0.613721</td>\n",
              "      <td>-0.760654</td>\n",
              "      <td>-0.371965</td>\n",
              "      <td>-0.753832</td>\n",
              "      <td>-0.451449</td>\n",
              "      <td>-0.303728</td>\n",
              "      <td>-1.043456</td>\n",
              "      <td>0.042248</td>\n",
              "      <td>2.474331</td>\n",
              "      <td>-1.389805</td>\n",
              "      <td>-0.323452</td>\n",
              "      <td>-0.131015</td>\n",
              "      <td>2.947365</td>\n",
              "      <td>0.106433</td>\n",
              "      <td>-0.375699</td>\n",
              "      <td>-0.817827</td>\n",
              "      <td>-0.365029</td>\n",
              "      <td>-0.736179</td>\n",
              "      <td>0.078568</td>\n",
              "      <td>-1.272535</td>\n",
              "      <td>-0.752311</td>\n",
              "      <td>1.063363</td>\n",
              "      <td>-0.399832</td>\n",
              "      <td>-0.864830</td>\n",
              "      <td>0.237418</td>\n",
              "      <td>-0.543032</td>\n",
              "      <td>-0.261251</td>\n",
              "      <td>-0.106171</td>\n",
              "      <td>-0.638665</td>\n",
              "      <td>-0.13702</td>\n",
              "      <td>-0.189721</td>\n",
              "      <td>0.272109</td>\n",
              "      <td>-0.274884</td>\n",
              "      <td>1.424805</td>\n",
              "      <td>-0.474512</td>\n",
              "      <td>-0.665942</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>7</th>\n",
              "      <td>716</td>\n",
              "      <td>716</td>\n",
              "      <td>NDARINV3ZRYG6UB</td>\n",
              "      <td>10</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",
              "      <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>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>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>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>...</td>\n",
              "      <td>0.633197</td>\n",
              "      <td>-0.333709</td>\n",
              "      <td>-0.139992</td>\n",
              "      <td>-0.540846</td>\n",
              "      <td>2.427284</td>\n",
              "      <td>1.152931</td>\n",
              "      <td>-0.168620</td>\n",
              "      <td>0.103277</td>\n",
              "      <td>0.857725</td>\n",
              "      <td>0.117351</td>\n",
              "      <td>0.327531</td>\n",
              "      <td>-0.398297</td>\n",
              "      <td>1.110883</td>\n",
              "      <td>-0.323452</td>\n",
              "      <td>-0.368455</td>\n",
              "      <td>-0.658629</td>\n",
              "      <td>-0.500448</td>\n",
              "      <td>-0.572905</td>\n",
              "      <td>1.057440</td>\n",
              "      <td>0.331896</td>\n",
              "      <td>-0.222763</td>\n",
              "      <td>1.181389</td>\n",
              "      <td>0.207931</td>\n",
              "      <td>0.176657</td>\n",
              "      <td>1.063363</td>\n",
              "      <td>-0.399832</td>\n",
              "      <td>-0.989884</td>\n",
              "      <td>0.131435</td>\n",
              "      <td>-0.000277</td>\n",
              "      <td>1.126744</td>\n",
              "      <td>-0.106171</td>\n",
              "      <td>1.058079</td>\n",
              "      <td>-0.13702</td>\n",
              "      <td>-0.189721</td>\n",
              "      <td>0.655672</td>\n",
              "      <td>-0.274884</td>\n",
              "      <td>0.082783</td>\n",
              "      <td>-0.474512</td>\n",
              "      <td>-0.553328</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>8</th>\n",
              "      <td>1467</td>\n",
              "      <td>1467</td>\n",
              "      <td>NDARINV8B8WV7MT</td>\n",
              "      <td>6</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</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>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>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>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",
              "      <td>0</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.266436</td>\n",
              "      <td>1.124449</td>\n",
              "      <td>-0.720845</td>\n",
              "      <td>0.658266</td>\n",
              "      <td>0.100951</td>\n",
              "      <td>0.824031</td>\n",
              "      <td>-0.760559</td>\n",
              "      <td>-0.451449</td>\n",
              "      <td>0.393144</td>\n",
              "      <td>2.610446</td>\n",
              "      <td>-0.908694</td>\n",
              "      <td>-0.863389</td>\n",
              "      <td>1.315134</td>\n",
              "      <td>-0.323452</td>\n",
              "      <td>-0.368455</td>\n",
              "      <td>0.142703</td>\n",
              "      <td>-0.191172</td>\n",
              "      <td>0.215921</td>\n",
              "      <td>-0.385460</td>\n",
              "      <td>-0.365029</td>\n",
              "      <td>-0.465960</td>\n",
              "      <td>-0.588253</td>\n",
              "      <td>0.980803</td>\n",
              "      <td>1.105625</td>\n",
              "      <td>-0.361203</td>\n",
              "      <td>-0.399832</td>\n",
              "      <td>0.482978</td>\n",
              "      <td>2.906629</td>\n",
              "      <td>-0.217379</td>\n",
              "      <td>-0.261251</td>\n",
              "      <td>-0.106171</td>\n",
              "      <td>-0.638665</td>\n",
              "      <td>-0.13702</td>\n",
              "      <td>-0.189721</td>\n",
              "      <td>-0.891805</td>\n",
              "      <td>-0.274884</td>\n",
              "      <td>-0.835985</td>\n",
              "      <td>-0.474512</td>\n",
              "      <td>-0.553328</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9</th>\n",
              "      <td>1022</td>\n",
              "      <td>1022</td>\n",
              "      <td>NDARINV5MHT9RR4</td>\n",
              "      <td>7</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>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>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>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>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>...</td>\n",
              "      <td>1.307923</td>\n",
              "      <td>0.157506</td>\n",
              "      <td>-0.139992</td>\n",
              "      <td>0.121646</td>\n",
              "      <td>-0.588333</td>\n",
              "      <td>0.046634</td>\n",
              "      <td>-0.256065</td>\n",
              "      <td>-0.451449</td>\n",
              "      <td>0.160853</td>\n",
              "      <td>-0.605515</td>\n",
              "      <td>2.752436</td>\n",
              "      <td>-0.726597</td>\n",
              "      <td>-0.226866</td>\n",
              "      <td>-0.323452</td>\n",
              "      <td>0.106424</td>\n",
              "      <td>-0.549357</td>\n",
              "      <td>-0.278703</td>\n",
              "      <td>0.531452</td>\n",
              "      <td>-0.568965</td>\n",
              "      <td>-0.365029</td>\n",
              "      <td>-0.161964</td>\n",
              "      <td>-0.890099</td>\n",
              "      <td>0.782363</td>\n",
              "      <td>-0.287827</td>\n",
              "      <td>-0.361203</td>\n",
              "      <td>-0.399832</td>\n",
              "      <td>0.802562</td>\n",
              "      <td>0.363028</td>\n",
              "      <td>1.845087</td>\n",
              "      <td>-0.261251</td>\n",
              "      <td>-0.106171</td>\n",
              "      <td>-0.638665</td>\n",
              "      <td>-0.13702</td>\n",
              "      <td>-0.189721</td>\n",
              "      <td>0.501365</td>\n",
              "      <td>-0.274884</td>\n",
              "      <td>0.671208</td>\n",
              "      <td>-0.474512</td>\n",
              "      <td>-0.215487</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>10 rows × 5020 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "   Unnamed: 0     X  ... con_R.IN_R.CER_count  kfold\n",
              "0        1799  1799  ...             0.572809      4\n",
              "1         876   876  ...            -0.778556      2\n",
              "2         783   783  ...            -0.665942      2\n",
              "3        1425  1425  ...             1.924174      3\n",
              "4         442   442  ...             4.739517      4\n",
              "5        1653  1653  ...             3.162925      2\n",
              "6        1316  1316  ...            -0.665942      0\n",
              "7         716   716  ...            -0.553328      2\n",
              "8        1467  1467  ...            -0.553328      2\n",
              "9        1022  1022  ...            -0.215487      0\n",
              "\n",
              "[10 rows x 5020 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 14
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AfYWzW_ZpSOk"
      },
      "source": [
        "# Dataset is ready! Let's define some important functions we need.\n",
        "\n",
        "We need to tune the hyperparameter when we train our model TabNet. We test some combinations of the hyperparameter and find the set of hyperparameters that gets the highest AUC performance.\n",
        "\n",
        "We want to extract important featuers, so we have to calculate the importance of the features."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "n52gHdH1czPl"
      },
      "source": [
        "## Define feature importance function to get the importance of the features"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9cWqc-nWd8nW"
      },
      "source": [
        "def feature(Num_feat, clf_res, test_data_processed, features):\n",
        "    importance_res = []\n",
        "    # Save and stack the k importances of the feautures in an array\n",
        "    for i in clf_res:\n",
        "        importance_clf =i.feature_importances_\n",
        "        importance_res.append(importance_clf)\n",
        "    \n",
        "    # Calculate the average of importance values ​​of features from each of k folds\n",
        "    importance=[importance_res[0][i]/5+importance_res[1][i]/5+ importance_res[2][i]/5+ importance_res[3][i]/5+ importance_res[4][i]/5 for i in range(len(importance_res[1]))]\n",
        "\n",
        "    feat_name_sort = test_data_processed[features].columns\n",
        "    important_features = pd.DataFrame([importance],columns = feat_name_sort, index=['Importance']) \n",
        "    important_features =important_features.transpose().sort_values(by=['Importance'], ascending=False)\n",
        "    important_features = important_features.head(Num_feat)\n",
        "\n",
        "    return important_features"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hG78GEgjejCz"
      },
      "source": [
        "## Define parameter-tuning function to find best hyperparameter"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bEI05yLReiKi"
      },
      "source": [
        "def find_bestpar(fold, train_data_processed, test_data_processed, features):\n",
        "    \n",
        "    \"\"\"Make test data\"\"\"\n",
        "    X_test = test_data_processed[features].values\n",
        "    Y_test = test_data_processed[target].values\n",
        "    \n",
        "    # Store maximum auc\n",
        "    max_auc= 0\n",
        "    # Store maximum hypterparameter set\n",
        "    max_hy = []\n",
        "\n",
        "    \"\"\"\n",
        "    # define hyperparameter space (original version)\n",
        "    n_ = [4,8,16]                              # \n",
        "    lr_ = [2e-2, 1e-2, 5e-3, 2e-3, 1e-3, 1e-4] # learning rate\n",
        "    w_ = [0.01, 0.001, 0.0001]                 # weight decay\n",
        "    g_ = [0.95, 0.99, 0.9]                     # scheduler params - gamma\n",
        "    ss_ = [10, 20, 30]                         # scheduler params - step_size\n",
        "    \"\"\"\n",
        "    # define hyperparameter space (quick version)\n",
        "    n_ = [4,16]\n",
        "    lr_ = [2e-2,1e-3]\n",
        "    w_ = [0.01,0.001]\n",
        "    g_ = [0.95,0.99]\n",
        "    ss_ = [10,30]\n",
        "    \n",
        "    \n",
        "    all_ = [n_, lr_, w_, g_, ss_]\n",
        "    h_space = [s for s in itertools.product(*all_)]\n",
        "    \n",
        "    # hyperparameter tuning\n",
        "    for hy in tqdm(h_space):\n",
        "        \n",
        "        \"\"\"define arrays to save the results of the validation & test\"\"\"\n",
        "        valid_res = []\n",
        "        test_auc_res = []\n",
        "        test_acc_res = []\n",
        "     \n",
        "        # cross-validation\n",
        "        for i in range(fold):\n",
        "            # define the model\n",
        "            clf = TabNetClassifier(n_a = hy[0],\n",
        "                                n_d = hy[0],\n",
        "                                optimizer_params = dict(lr=hy[1], weight_decay=hy[2]),\n",
        "                                scheduler_params={\"step_size\":hy[4], \"gamma\":hy[3]},\n",
        "                                scheduler_fn=torch.optim.lr_scheduler.StepLR,\n",
        "                                verbose=0)\n",
        "            \n",
        "            df_train = train_data_processed[train_data_processed.kfold != i]  # train : 4/5\n",
        "            df_valid = train_data_processed[train_data_processed.kfold == i]  # validation : 1/5\n",
        "            \n",
        "            X_train = df_train[features].values\n",
        "            Y_train = df_train[target].values\n",
        "            \n",
        "            X_valid = df_valid[features].values\n",
        "            Y_valid = df_valid[target].values\n",
        "            \n",
        "            # train the model with our train dataset\n",
        "            clf.fit(X_train, Y_train, eval_set=[(X_train, Y_train), (X_valid, Y_valid)], \n",
        "                    eval_name=['train', 'valid'], eval_metric=['auc'],\n",
        "                    max_epochs=200 , patience=20)\n",
        "       \n",
        "            # test the model which is trained above\n",
        "            preds_acc = clf.predict(X_test)\n",
        "            preds_prob = clf.predict_proba(X_test)\n",
        "            test_auc = roc_auc_score(y_score=preds_prob[:,1], y_true=Y_test)\n",
        "            test_acc = accuracy_score(preds_acc, Y_test)\n",
        "            \n",
        "            valid_res.append(clf.best_cost)\n",
        "            test_auc_res.append(test_auc)\n",
        "            test_acc_res.append(test_acc)\n",
        "            print('[%3d/%4d] '%(i+1, fold),'Valid score: %2f'% clf.best_cost, 'Test AUC: %.3f%%'%test_auc, 'Test ACC: %.3f%%'%test_acc)\n",
        "    \n",
        "        \"\"\"print average and standard deviation of validation and testing\"\"\"\n",
        "        print(\"=====parameter valid, test score=====\")\n",
        "        print(\"Validation average: %3f\"%np.mean(valid_res), \"Test AUC average: %3f\"%np.mean(test_auc_res), \"Test ACC average: %3f\"%np.mean(test_acc_res))\n",
        "\n",
        "        # save best AUC performance and the hyperparameter set that make the best AUC score\n",
        "        if np.mean(test_auc_res)>max_auc:\n",
        "            print(\"Find new maximum AUC!!\")\n",
        "            max_hy = hy\n",
        "            max_auc = np.mean(test_auc_res)\n",
        "    \n",
        "    return max_hy"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1DoJJRmQfBFj"
      },
      "source": [
        "## Define Training function with best parameter"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XbU5TgO3fGa8"
      },
      "source": [
        "def bestpar_tuning(fold, train_data_processed, test_data_processed, max_hy, features):\n",
        "    # bring the hyperparameter set making the best performance\n",
        "    hy = max_hy\n",
        "    print(\"Max hy:\" ,hy)\n",
        "\n",
        "    X_test = test_data_processed[features].values\n",
        "    Y_test = test_data_processed[target].values\n",
        "\n",
        "    \"\"\"define arrays saving the results of validation & test\"\"\"\n",
        "    valid_res = []\n",
        "    test_auc_res = []\n",
        "    test_acc_res = []\n",
        "    clf_res = []\n",
        "    preds_prob_res = []\n",
        "     \n",
        "    y_valid_true = []\n",
        "    y_valid_pred = []\n",
        "    y_test_pred = []\n",
        "    \n",
        "    y_valid_subject = []\n",
        "    y_test_subject = []\n",
        "    \n",
        "    \n",
        "    for i in range(fold):\n",
        "        clf = TabNetClassifier(n_a = hy[0],n_d = hy[0],\n",
        "                           optimizer_params = dict(lr=hy[1], weight_decay=hy[2]),\n",
        "                           scheduler_params={\"step_size\":hy[4], \"gamma\":hy[3]},\n",
        "                           scheduler_fn=torch.optim.lr_scheduler.StepLR,\n",
        "                           verbose=0)\n",
        "        \n",
        "        df_train = train_data_processed[train_data_processed.kfold != i]  # using the data of 4 folds dataset as training dataset\n",
        "        df_valid = train_data_processed[train_data_processed.kfold == i]  # using the data of 1 fold dataset as validation dataset\n",
        "            \n",
        "        X_train = df_train[features].values\n",
        "        Y_train = df_train[target].values\n",
        "            \n",
        "        X_valid = df_valid[features].values\n",
        "        Y_valid = df_valid[target].values\n",
        "        \n",
        "        # bring subject key before the training\n",
        "        y_valid_subject.append(df_valid['subjectkey'].values)\n",
        "        y_test_subject.append(test_data_processed['subjectkey'].values)  \n",
        "        y_valid_true.append(Y_valid)        \n",
        "        \n",
        "        # Training\n",
        "        clf.fit(X_train, Y_train, eval_set=[(X_train, Y_train), (X_valid, Y_valid)], \n",
        "                        eval_name=['train', 'valid'], eval_metric=['auc'],\n",
        "                        max_epochs=200 , patience=20)\n",
        "        \n",
        "        # Results\n",
        "        preds_prob_val= clf.predict_proba(X_valid)[:,1]\n",
        "        y_valid_pred.append(preds_prob_val)\n",
        "        y_test_pred.append(clf.predict(X_test))\n",
        "\n",
        "        preds_acc = clf.predict(X_test)\n",
        "        preds_prob = clf.predict_proba(X_test)\n",
        "        test_auc = roc_auc_score(y_score=preds_prob[:,1], y_true=Y_test)\n",
        "        test_acc = accuracy_score(preds_acc, Y_test)\n",
        "                    \n",
        "        valid_res.append(clf.best_cost)\n",
        "        test_auc_res.append(test_auc)\n",
        "        test_acc_res.append(test_acc)\n",
        "        preds_prob_res.append(preds_prob)\n",
        "        clf_res.append(clf)\n",
        "\n",
        "        print('[%3d/%4d] '%(i+1, fold),'Valid score: %2f'% clf.best_cost, 'Test AUC: %.3f'%test_auc, 'Test ACC: %.3f'%test_acc)\n",
        "\n",
        "    \"\"\"print average and standard deviation of validation and testing\"\"\"\n",
        "    print(\"Validation average: %3f\"%np.mean(valid_res), \"Test AUC average: %3f \"%np.mean(test_auc_res), \"Test ACC average: %3f\"%np.mean(test_acc_res))\n",
        "    \n",
        "    preds_prob=[preds_prob_res[0][i]/5+preds_prob_res[1][i]/5+ preds_prob_res[2][i]/5+ preds_prob_res[3][i]/5+ preds_prob_res[4][i]/5 for i in range(len(preds_prob_res[1]))]\n",
        "    preds_prob = np.array(preds_prob)\n",
        "    valid_result = np.mean(valid_res)\n",
        "    test_auc = np.mean(test_auc_res)\n",
        "    test_acc = np.mean(test_acc_res) \n",
        "    \n",
        "    #return test_auc,test_acc ,valid_result, clf_res, preds_prob, X_test, Y_test, y_valid_pred, y_test_pred, y_valid_subject, y_test_subject\n",
        "    return test_auc,test_acc ,valid_result, clf_res, preds_prob, X_test, Y_test, y_valid_pred, y_test_pred, y_valid_subject, y_test_subject, y_valid_true\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "T9Q2XL4dfIzY"
      },
      "source": [
        "## Define run function\n",
        "\n",
        "\n",
        "*   Call the functions defined above.\n",
        "*   Get and return the results including test AUC, ACC score.\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fcoCJfG7fZHF"
      },
      "source": [
        "def run(train_data_processed, test_data_processed, fold, Num_feat, features):\n",
        "    name_test = test_data_processed['subjectkey'].values\n",
        "    \n",
        "    # Finding best performance hyperparameter (Hyperparameter tuning)\n",
        "    max_hy = find_bestpar(fold, train_data_processed, test_data_processed, features)\n",
        "    \n",
        "    # Training with best hyperparameter\n",
        "    test_auc,test_acc ,valid_result, clf_res, preds_prob, X_test, Y_test, y_valid_pred, y_test_pred, y_valid_subject, y_test_subject, y_valid_true= bestpar_tuning(fold, \n",
        "                                                                                        train_data_processed, \n",
        "                                                                                        test_data_processed, \n",
        "                                                                                        max_hy, \n",
        "                                                                                        features)\n",
        "    \n",
        "    # Calculating the importances of the features and extracting 50 important features\n",
        "    import_feat=feature(Num_feat, clf_res, test_data_processed, features)\n",
        "    \n",
        "    return test_auc,test_acc ,valid_result, clf_res, preds_prob, X_test, Y_test, import_feat, name_test, y_valid_pred, y_test_pred, y_valid_subject, y_test_subject, y_valid_true\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "m3NrdSA1yiSv"
      },
      "source": [
        "## Define function saving the results as csv file\n",
        "\n",
        "This is used for stacking ensemble model."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ORdz1PXwqhh-"
      },
      "source": [
        "def save_prob_with_true(model):\n",
        "    modeltype= \"\" \n",
        "    combined_model=pd.DataFrame({f\"subjectkey\": model.y_test_subject[0], f\"Y_{modeltype}\":model.Y_test, f\"preds_prob_{modeltype}\" :model.preds_prob[:,1]} )\n",
        "    combined_model.to_csv(f\"combined_forROC_{modeltype}_.csv\")\n",
        "\n",
        "    return combined_model\n",
        "\n",
        "def save_prob_with_true_valid(model):\n",
        "    modeltype= \"\" \n",
        "    \n",
        "    combined_model=pd.DataFrame({f\"subjectkey\": list(itertools.chain(*model.y_valid_subject)), f\"Y_{modeltype}\":list(itertools.chain(*model.y_valid_true)), f\"preds_prob_{modeltype}\" :list(itertools.chain(*model.y_valid_pred) )})\n",
        "    combined_model.to_csv(f\"combined_forROC_{modeltype}_.csv\")\n",
        "\n",
        "    return combined_model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XDXjUBW5qNJz"
      },
      "source": [
        "## Define main function\n",
        "\n",
        "Call the run function with our preprocessed train & test dataset.\n",
        "\n",
        "Save the results returned for confirm the results."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qrhZ-y5qqTiG"
      },
      "source": [
        "class model():\n",
        "    def __init__(self, train_data_processed, test_data_processed, Num_FOLDS, Num_feat, features):\n",
        "        test_auc,test_acc ,valid_result, clf_res, preds_prob, X_test, Y_test, import_feat, name_test, y_valid_pred, y_test_pred, y_valid_subject, y_test_subject, y_valid_true = run(train_data_processed,\n",
        "                                                                              test_data_processed,\n",
        "                                                                              Num_FOLDS, \n",
        "                                                                              Num_feat, \n",
        "                                                                              features)\n",
        "    \n",
        "    \n",
        "        self.train_data_processed = train_data_processed\n",
        "        self.test_auc = test_auc\n",
        "        self.test_acc = test_acc\n",
        "        self.valid_result = valid_result\n",
        "        self.clf_res = clf_res \n",
        "        self.preds_prob = preds_prob \n",
        "        self.X_test = X_test\n",
        "        self.Y_test = Y_test\n",
        "        self.import_feat =  import_feat\n",
        "        self.name_test = name_test\n",
        "        self.features = features\n",
        "        self.y_valid_pred =y_valid_pred\n",
        "        self.y_test_pred = y_test_pred\n",
        "        self.y_valid_subject= y_valid_subject\n",
        "        self.y_test_subject = y_test_subject\n",
        "        self.y_valid_true = y_valid_true\n",
        "\n",
        "        test_prob_result = save_prob_with_true(self)\n",
        "        valid_prob_result = save_prob_with_true_valid(self)\n",
        "        self.test_prob_result= test_prob_result\n",
        "        self.valid_prob_result = valid_prob_result        "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2-MyLei-qSy-"
      },
      "source": [
        "# All dataset and functions are ready! Let's train our model. "
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Select the features we want to use\n",
        "features = [col for col in train_data_processed.columns if col in psychosocial + rsf_mri + structural_mri + diffusion_mri]\n",
        "\n",
        "# Call the model function to tune the hyperparameter and train the TabNet model\n",
        "rsf_s_d_model = model(train_data_processed, test_data_processed, Num_FOLDS, Num_feat, features)"
      ],
      "metadata": {
        "id": "VrI3PW8CsBmk"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oqNA5I9E4QhF"
      },
      "source": [
        "# We are done! Let's get the results"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "FmhI2QJx1EVV",
        "outputId": "0949a1b4-6cd9-4cfa-fcfd-e06fabb94214"
      },
      "source": [
        "print(rsf_s_d_model.valid_result)\n",
        "print(rsf_s_d_model.test_auc)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0.7492064576989276\n",
            "0.6568505692221958\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "zOkz3DsBztBD",
        "outputId": "f2290dd6-f610-4a60-cacf-d66bd8fed92f"
      },
      "source": [
        "rsf_s_d_model.import_feat"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>Importance</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>Internal.and.External.Comobidity</th>\n",
              "      <td>0.166500</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Depress.CBCL</th>\n",
              "      <td>0.042368</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>WithDep.CBCL</th>\n",
              "      <td>0.031361</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>fes_q1_1</th>\n",
              "      <td>0.029827</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>rh_lateralorbitofrontal_thickness._.6</th>\n",
              "      <td>0.029299</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>External.CBCL</th>\n",
              "      <td>0.022822</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>rh_G.S_frontomargin_area._.14</th>\n",
              "      <td>0.020294</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>AnxDep.CBCL</th>\n",
              "      <td>0.019561</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>rh_S_parieto_occipital_volume._.17</th>\n",
              "      <td>0.017230</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>rsfmri_cor_network.gordon_auditory_subcort.aseg_accumbens.area.rh</th>\n",
              "      <td>0.013686</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.AM_R.PCG_count</th>\n",
              "      <td>0.012798</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>AnxDisord.CBCL</th>\n",
              "      <td>0.012727</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.LOFG_R.CACG_count</th>\n",
              "      <td>0.012335</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_R.PU_R.CACG_count</th>\n",
              "      <td>0.012010</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Internal.CBCL</th>\n",
              "      <td>0.011474</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.PCU_L.SMG_count</th>\n",
              "      <td>0.011126</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.CACG_R.PCG_count</th>\n",
              "      <td>0.010758</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>rsfmri_cor_network.gordon_smmouth_network.gordon_retrosplenialtemporal</th>\n",
              "      <td>0.010615</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Sluggish.Cognitive.Tempo..SCT..CBCL</th>\n",
              "      <td>0.009972</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.FG_R.SFG_count</th>\n",
              "      <td>0.009949</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_R.LG_R.PCAL_count</th>\n",
              "      <td>0.009714</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.PCU_R.TTG_count</th>\n",
              "      <td>0.009211</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_R.FG_R.SPG_count</th>\n",
              "      <td>0.009091</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.TP_L.AM_count</th>\n",
              "      <td>0.008990</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.CMFG_L.POP_count</th>\n",
              "      <td>0.008717</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.IPG_L.ICG_count</th>\n",
              "      <td>0.008376</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.PCG_L.PCU_count</th>\n",
              "      <td>0.007775</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>rh_S_circular_insula_sup_volume._.17</th>\n",
              "      <td>0.007771</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.SPG_R.POP_count</th>\n",
              "      <td>0.007618</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_R.EC_R.LG_count</th>\n",
              "      <td>0.007596</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>rsfmri_cor_network.gordon_smhand_subcort.aseg_pallidum.lh</th>\n",
              "      <td>0.007558</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.CER_R.AC_count</th>\n",
              "      <td>0.007124</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>rsfmri_cor_network.gordon_default_subcort.aseg_cerebellum.cortex.rh</th>\n",
              "      <td>0.007053</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_R.HI_R.AC_count</th>\n",
              "      <td>0.006920</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.AM_R.PaCG_count</th>\n",
              "      <td>0.006808</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.POR_L.RACG_count</th>\n",
              "      <td>0.006621</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.PoCG_R.SMG_count</th>\n",
              "      <td>0.006514</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>fes_q2_1</th>\n",
              "      <td>0.006180</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.LOG_R.LOFG_count</th>\n",
              "      <td>0.005900</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.POR_R.BSTS_count</th>\n",
              "      <td>0.005656</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.PA_R.BSTS_count</th>\n",
              "      <td>0.005521</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>fes_q4_2</th>\n",
              "      <td>0.005511</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>rh_isthmuscingulate_thickness._.6</th>\n",
              "      <td>0.005490</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Aggressive.CBCL</th>\n",
              "      <td>0.005391</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.PaCG_R.PTR_count</th>\n",
              "      <td>0.005387</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.CA_R.AM_count</th>\n",
              "      <td>0.005380</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>rh_precentral_thickness._.6</th>\n",
              "      <td>0.005330</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_R.PCG_R.TP_count</th>\n",
              "      <td>0.005289</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>lh_Lat_Fis.ant.Horizont_volume._.13</th>\n",
              "      <td>0.005105</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>con_L.PoCG_R.RACG_count</th>\n",
              "      <td>0.004971</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                                                    Importance\n",
              "Internal.and.External.Comobidity                      0.166500\n",
              "Depress.CBCL                                          0.042368\n",
              "WithDep.CBCL                                          0.031361\n",
              "fes_q1_1                                              0.029827\n",
              "rh_lateralorbitofrontal_thickness._.6                 0.029299\n",
              "External.CBCL                                         0.022822\n",
              "rh_G.S_frontomargin_area._.14                         0.020294\n",
              "AnxDep.CBCL                                           0.019561\n",
              "rh_S_parieto_occipital_volume._.17                    0.017230\n",
              "rsfmri_cor_network.gordon_auditory_subcort.aseg...    0.013686\n",
              "con_L.AM_R.PCG_count                                  0.012798\n",
              "AnxDisord.CBCL                                        0.012727\n",
              "con_L.LOFG_R.CACG_count                               0.012335\n",
              "con_R.PU_R.CACG_count                                 0.012010\n",
              "Internal.CBCL                                         0.011474\n",
              "con_L.PCU_L.SMG_count                                 0.011126\n",
              "con_L.CACG_R.PCG_count                                0.010758\n",
              "rsfmri_cor_network.gordon_smmouth_network.gordo...    0.010615\n",
              "Sluggish.Cognitive.Tempo..SCT..CBCL                   0.009972\n",
              "con_L.FG_R.SFG_count                                  0.009949\n",
              "con_R.LG_R.PCAL_count                                 0.009714\n",
              "con_L.PCU_R.TTG_count                                 0.009211\n",
              "con_R.FG_R.SPG_count                                  0.009091\n",
              "con_L.TP_L.AM_count                                   0.008990\n",
              "con_L.CMFG_L.POP_count                                0.008717\n",
              "con_L.IPG_L.ICG_count                                 0.008376\n",
              "con_L.PCG_L.PCU_count                                 0.007775\n",
              "rh_S_circular_insula_sup_volume._.17                  0.007771\n",
              "con_L.SPG_R.POP_count                                 0.007618\n",
              "con_R.EC_R.LG_count                                   0.007596\n",
              "rsfmri_cor_network.gordon_smhand_subcort.aseg_p...    0.007558\n",
              "con_L.CER_R.AC_count                                  0.007124\n",
              "rsfmri_cor_network.gordon_default_subcort.aseg_...    0.007053\n",
              "con_R.HI_R.AC_count                                   0.006920\n",
              "con_L.AM_R.PaCG_count                                 0.006808\n",
              "con_L.POR_L.RACG_count                                0.006621\n",
              "con_L.PoCG_R.SMG_count                                0.006514\n",
              "fes_q2_1                                              0.006180\n",
              "con_L.LOG_R.LOFG_count                                0.005900\n",
              "con_L.POR_R.BSTS_count                                0.005656\n",
              "con_L.PA_R.BSTS_count                                 0.005521\n",
              "fes_q4_2                                              0.005511\n",
              "rh_isthmuscingulate_thickness._.6                     0.005490\n",
              "Aggressive.CBCL                                       0.005391\n",
              "con_L.PaCG_R.PTR_count                                0.005387\n",
              "con_L.CA_R.AM_count                                   0.005380\n",
              "rh_precentral_thickness._.6                           0.005330\n",
              "con_R.PCG_R.TP_count                                  0.005289\n",
              "lh_Lat_Fis.ant.Horizont_volume._.13                   0.005105\n",
              "con_L.PoCG_R.RACG_count                               0.004971"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 24
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7OfoiaO91zbM"
      },
      "source": [
        "## Draw ROC curve graph\n",
        "1. Define the function drawing the roc curve"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eOdfQHi01LM7"
      },
      "source": [
        "def rocvis(true , prob , label ) :\n",
        "    from sklearn.metrics import roc_curve\n",
        "    if type(true[0]) == str :\n",
        "        from sklearn.preprocessing import LabelEncoder\n",
        "        le = LabelEncoder()\n",
        "        true = le.fit_transform(true)\n",
        "    else :\n",
        "        pass\n",
        "    fpr, tpr, thresholds = roc_curve(true, prob)\n",
        "    plt.plot(fpr, tpr, marker='.', label = label  )"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "YPr9M2Jl1943"
      },
      "source": [
        "2. Draw the roc curve of our model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 349
        },
        "id": "hk-VrmMg1M7N",
        "outputId": "632f4bf5-1e0e-45cb-8e2b-62464b62981d"
      },
      "source": [
        "fig , ax = plt.subplots(figsize= (20,10))\n",
        "plt.plot([0, 1], [0, 1], linestyle='--')\n",
        "rocvis(rsf_s_d_model.Y_test , rsf_s_d_model.preds_prob[:,1] , f\"rs_revise:{round(rsf_s_d_model.test_auc*100, 2)}%\")\n",
        "\n",
        "\n",
        "dataset_name=\"rsf+s+d\"\n",
        "\n",
        "plt.legend(fontsize = 40)\n",
        "plt.title(dataset_name, fontsize= 50)\n",
        "plt.xlabel(\"FP rate\", fontsize =30)\n",
        "plt.ylabel(\"TP rate\", fontsize =30)\n",
        "plt.xticks(size = 30)\n",
        "plt.yticks(size = 30)\n",
        "\n",
        "plt.tight_layout()\n",
        "plt.savefig(f'{dataset_name}_ROC.png')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1440x720 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Hyng6rNA47Ea"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}