src/GBC.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "source": [
        "## Set up the environment and data folder"
      ],
      "metadata": {
        "id": "ih2ZFO4BZA92"
      },
      "id": "ih2ZFO4BZA92"
    },
    {
      "cell_type": "code",
      "source": [
        "## Download the dataset from github repo\n",
        "!wget https://media.githubusercontent.com/media/SIGCOMM22-5GMobility/artifact/master/data/table3-system_evaluation/D1.csv\n",
        "!wget https://media.githubusercontent.com/media/SIGCOMM22-5GMobility/artifact/master/data/table3-system_evaluation/D2.csv"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7tFhXXfHX3Hd",
        "outputId": "58dd7224-622e-4133-d69c-87e2b209d240"
      },
      "id": "7tFhXXfHX3Hd",
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "--2022-06-26 01:15:39--  https://media.githubusercontent.com/media/SIGCOMM22-5GMobility/artifact/master/data/table3-system_evaluation/D1.csv\n",
            "Resolving media.githubusercontent.com (media.githubusercontent.com)... 185.199.108.133, 185.199.109.133, 185.199.110.133, ...\n",
            "Connecting to media.githubusercontent.com (media.githubusercontent.com)|185.199.108.133|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 345077566 (329M) [text/plain]\n",
            "Saving to: ‘D1.csv.2’\n",
            "\n",
            "D1.csv.2            100%[===================>] 329.09M   205MB/s    in 1.6s    \n",
            "\n",
            "2022-06-26 01:15:45 (205 MB/s) - ‘D1.csv.2’ saved [345077566/345077566]\n",
            "\n",
            "--2022-06-26 01:15:45--  https://media.githubusercontent.com/media/SIGCOMM22-5GMobility/artifact/master/data/table3-system_evaluation/D2.csv\n",
            "Resolving media.githubusercontent.com (media.githubusercontent.com)... 185.199.108.133, 185.199.109.133, 185.199.110.133, ...\n",
            "Connecting to media.githubusercontent.com (media.githubusercontent.com)|185.199.108.133|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 293077812 (280M) [text/plain]\n",
            "Saving to: ‘D2.csv.2’\n",
            "\n",
            "D2.csv.2            100%[===================>] 279.50M   212MB/s    in 1.3s    \n",
            "\n",
            "2022-06-26 01:16:04 (212 MB/s) - ‘D2.csv.2’ saved [293077812/293077812]\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "## configure directory where data is stored\n",
        "import os\n",
        "DATA_FOLDER = os.path.curdir"
      ],
      "metadata": {
        "id": "4VvDn4MfX32e"
      },
      "id": "4VvDn4MfX32e",
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "id": "considered-component",
      "metadata": {
        "id": "considered-component"
      },
      "source": [
        "## Import packages"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "id": "applied-payment",
      "metadata": {
        "id": "applied-payment"
      },
      "outputs": [],
      "source": [
        "#!/usr/bin/env python3\n",
        "import random\n",
        "import pprint\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "from os import path\n",
        "from scipy import stats\n",
        "import matplotlib.pyplot as plt\n",
        "import seaborn as sn\n",
        "from sklearn import preprocessing\n",
        "from sklearn import tree\n",
        "from sklearn.svm import SVC\n",
        "from sklearn.linear_model import LinearRegression\n",
        "from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier, GradientBoostingClassifier\n",
        "from sklearn.metrics import confusion_matrix, f1_score, recall_score, precision_score, accuracy_score, plot_roc_curve\n",
        "\n",
        "# from utils.context import data_dir_shared_pred\n",
        "\n",
        "np.set_printoptions(suppress=True)\n",
        "pd.set_option('display.float_format',lambda x : '%.3f' % x)"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Configure the parameters for data processing and training\n",
        "\n",
        "```Split method```:\n",
        "Split the whole database into split 1 and split 2 based on the ratio. Do the negative sampling on split 1 for training. Test the model on split 2."
      ],
      "metadata": {
        "id": "Lt7kAMOjgG13"
      },
      "id": "Lt7kAMOjgG13"
    },
    {
      "cell_type": "code",
      "source": [
        "####################################\n",
        "### Config\n",
        "####################################\n",
        "PARAMETERS = {}\n",
        "\n",
        "# Read Database\n",
        "PARAMETERS[\"data_folder\"] = DATA_FOLDER\n",
        "PARAMETERS[\"file_name\"] = \"D1\"\n",
        "# PARAMETERS[\"file_name\"] = \"D2\"\n",
        "\n",
        "# Task\n",
        "PARAMETERS[\"label\"] = \"all_ho_category\"\n",
        "PARAMETERS[\"binary\"] = True\n",
        "PARAMETERS[\"feature_list\"] = [\"lon\", \"lat\", 'speed', 'lte_serv_rsrp', 'lte_serv_rsrq', 'lte_serv_rssi','lte_nbr_rsrp', 'lte_nbr_rsrq', 'lte_cqi0', 'lte_cqi1', 'dl_tput', \\\n",
        "                              'nr_serv_rsrp', 'nr_serv_rsrq', 'nr_serv_sinr', 'nr_cqi', 'nr_nbr_rsrp', 'nr_nbr_rsrq']  # , 'irat_rsrp', 'irat_rsrq'\n",
        "\n",
        "# Process the dataset \n",
        "PARAMETERS[\"split_mode\"] = \"train_neg_sampling\"\n",
        "\n",
        "####################################\n",
        "### Config (Hyper-parameters) \n",
        "####################################\n",
        "# Read Database\n",
        "PARAMETERS[\"sample_rate\"] = 50  # ms \n",
        "\n",
        "# Process the dataset \n",
        "PARAMETERS[\"train_ratio\"] = 0.7\n",
        "PARAMETERS[\"hist_window\"] = 20\n",
        "PARAMETERS[\"pred_window\"] = 20\n",
        "if PARAMETERS[\"file_name\"] == \"D1\":\n",
        "  PARAMETERS[\"neg_size\"] = 10000\n",
        "elif PARAMETERS[\"file_name\"] == \"D2\":\n",
        "  PARAMETERS[\"neg_size\"] = 50000\n",
        "\n",
        "# Machine Learning \n",
        "PARAMETERS[\"batch_size\"] = 64\n",
        "PARAMETERS[\"epochs\"] = 200\n",
        "PARAMETERS[\"latent_dim\"] = 32"
      ],
      "metadata": {
        "id": "JUhq4c_sY5tY"
      },
      "id": "JUhq4c_sY5tY",
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Preprocess the data\n",
        "It will take a couple of minutes (D2 will take 15min in Colab)."
      ],
      "metadata": {
        "id": "AZAyl5wqbm7v"
      },
      "id": "AZAyl5wqbm7v"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "id": "verified-angel",
      "metadata": {
        "id": "verified-angel",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 626
        },
        "outputId": "585a5b6a-f22e-4f38-9ef7-16c33e4e157e"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Index(['TIME_STAMP', 'lon', 'lat', 'speed', 'technology', 'lte_band',\n",
            "       'lte_serv_rsrp', 'lte_serv_rsrq', 'lte_serv_rssi', 'lte_cqi0',\n",
            "       'lte_cqi1', 'nr_band', 'nr_serv_rsrp', 'nr_serv_rsrq', 'nr_serv_sinr',\n",
            "       'nr_cqi', 'dl_tput', 'ul_tput', 'lte_nbr_rsrp', 'lte_nbr_rsrq',\n",
            "       'irat_rsrp', 'irat_rsrq', 'nr_nbr_rsrp', 'nr_nbr_rsrq',\n",
            "       'lte_ho_category', 'nr_ho_category', 'OLD_TIME_STAMP',\n",
            "       'all_ho_category', 'r_time'],\n",
            "      dtype='object')\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "          r_time     lon    lat  speed  lte_serv_rsrp  lte_serv_rsrq  \\\n",
              "0              0 -71.053 42.367  3.000        -76.240        -10.016   \n",
              "1             50 -71.053 42.367  3.000        -76.393         -9.599   \n",
              "2            100 -71.053 42.367  3.000        -77.722        -10.073   \n",
              "3            150 -71.053 42.367  3.000        -75.928         -9.228   \n",
              "4            200 -71.053 42.367  3.000        -75.340         -9.689   \n",
              "...          ...     ...    ...    ...            ...            ...   \n",
              "317855  15892750 -71.053 42.367  0.000        -69.242         -8.337   \n",
              "317856  15892800 -71.053 42.367  0.000        -69.242         -8.337   \n",
              "317857  15892850 -71.053 42.367  0.000        -69.242         -8.337   \n",
              "317858  15892900 -71.053 42.367  0.000        -68.890         -8.481   \n",
              "317859  15892950 -71.053 42.367  0.000        -68.890         -8.481   \n",
              "\n",
              "        lte_serv_rssi  lte_cqi0  lte_cqi1  nr_serv_rsrp  ...  ul_tput  \\\n",
              "0             -46.430    14.000     0.000       -89.630  ...    2.034   \n",
              "1             -47.974    14.000     0.000       -89.630  ...    2.034   \n",
              "2             -48.718    14.000     0.000       -90.310  ...    2.034   \n",
              "3             -44.980    14.000     0.000       -90.310  ...    2.214   \n",
              "4             -45.404    11.429     0.000       -90.310  ...    2.250   \n",
              "...               ...       ...       ...           ...  ...      ...   \n",
              "317855        -39.939     7.381     8.571       -86.680  ...    0.000   \n",
              "317856        -39.939     7.381     8.571       -86.680  ...    0.000   \n",
              "317857        -39.939     7.381     8.571       -86.680  ...    0.000   \n",
              "317858        -39.912    10.048     5.571       -86.680  ...    0.000   \n",
              "317859        -39.912    10.048     5.571       -86.680  ...    0.000   \n",
              "\n",
              "        lte_nbr_rsrp  lte_nbr_rsrq  irat_rsrp  irat_rsrq  nr_nbr_rsrp  \\\n",
              "0            -79.940       -15.630    -96.010    -10.510     -117.330   \n",
              "1            -79.940       -15.630    -96.010    -10.510     -117.330   \n",
              "2            -79.940       -15.630    -96.010    -10.510     -116.840   \n",
              "3            -80.420       -14.273    -96.010    -10.510     -116.840   \n",
              "4            -81.380       -11.560    -96.010    -10.510     -116.840   \n",
              "...              ...           ...        ...        ...          ...   \n",
              "317855       -80.758       -17.966    -99.500    -10.510     -109.700   \n",
              "317856       -80.758       -17.966    -99.500    -10.510     -109.700   \n",
              "317857       -80.758       -17.966    -99.500    -10.510     -109.700   \n",
              "317858       -80.750       -16.001    -99.500    -10.510     -109.700   \n",
              "317859       -80.750       -16.001    -99.500    -10.510     -109.700   \n",
              "\n",
              "        nr_nbr_rsrq  lte_ho_category  nr_ho_category  all_ho_category  \n",
              "0           -16.590            0.000           0.000            0.000  \n",
              "1           -16.590            0.000           0.000            0.000  \n",
              "2           -16.050            0.000           0.000            0.000  \n",
              "3           -16.050            0.000           0.000            0.000  \n",
              "4           -16.050            0.000           0.000            0.000  \n",
              "...             ...              ...             ...              ...  \n",
              "317855      -12.280            0.000           0.000            0.000  \n",
              "317856      -12.280            0.000           0.000            0.000  \n",
              "317857      -12.280            0.000           0.000            0.000  \n",
              "317858      -12.280            0.000           0.000            0.000  \n",
              "317859      -12.280            0.000           0.000            0.000  \n",
              "\n",
              "[317860 rows x 24 columns]"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-4382df32-7f26-4bbe-a23f-5989289e9038\">\n",
              "    <div class=\"colab-df-container\">\n",
              "      <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>r_time</th>\n",
              "      <th>lon</th>\n",
              "      <th>lat</th>\n",
              "      <th>speed</th>\n",
              "      <th>lte_serv_rsrp</th>\n",
              "      <th>lte_serv_rsrq</th>\n",
              "      <th>lte_serv_rssi</th>\n",
              "      <th>lte_cqi0</th>\n",
              "      <th>lte_cqi1</th>\n",
              "      <th>nr_serv_rsrp</th>\n",
              "      <th>...</th>\n",
              "      <th>ul_tput</th>\n",
              "      <th>lte_nbr_rsrp</th>\n",
              "      <th>lte_nbr_rsrq</th>\n",
              "      <th>irat_rsrp</th>\n",
              "      <th>irat_rsrq</th>\n",
              "      <th>nr_nbr_rsrp</th>\n",
              "      <th>nr_nbr_rsrq</th>\n",
              "      <th>lte_ho_category</th>\n",
              "      <th>nr_ho_category</th>\n",
              "      <th>all_ho_category</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>0</td>\n",
              "      <td>-71.053</td>\n",
              "      <td>42.367</td>\n",
              "      <td>3.000</td>\n",
              "      <td>-76.240</td>\n",
              "      <td>-10.016</td>\n",
              "      <td>-46.430</td>\n",
              "      <td>14.000</td>\n",
              "      <td>0.000</td>\n",
              "      <td>-89.630</td>\n",
              "      <td>...</td>\n",
              "      <td>2.034</td>\n",
              "      <td>-79.940</td>\n",
              "      <td>-15.630</td>\n",
              "      <td>-96.010</td>\n",
              "      <td>-10.510</td>\n",
              "      <td>-117.330</td>\n",
              "      <td>-16.590</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>50</td>\n",
              "      <td>-71.053</td>\n",
              "      <td>42.367</td>\n",
              "      <td>3.000</td>\n",
              "      <td>-76.393</td>\n",
              "      <td>-9.599</td>\n",
              "      <td>-47.974</td>\n",
              "      <td>14.000</td>\n",
              "      <td>0.000</td>\n",
              "      <td>-89.630</td>\n",
              "      <td>...</td>\n",
              "      <td>2.034</td>\n",
              "      <td>-79.940</td>\n",
              "      <td>-15.630</td>\n",
              "      <td>-96.010</td>\n",
              "      <td>-10.510</td>\n",
              "      <td>-117.330</td>\n",
              "      <td>-16.590</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>100</td>\n",
              "      <td>-71.053</td>\n",
              "      <td>42.367</td>\n",
              "      <td>3.000</td>\n",
              "      <td>-77.722</td>\n",
              "      <td>-10.073</td>\n",
              "      <td>-48.718</td>\n",
              "      <td>14.000</td>\n",
              "      <td>0.000</td>\n",
              "      <td>-90.310</td>\n",
              "      <td>...</td>\n",
              "      <td>2.034</td>\n",
              "      <td>-79.940</td>\n",
              "      <td>-15.630</td>\n",
              "      <td>-96.010</td>\n",
              "      <td>-10.510</td>\n",
              "      <td>-116.840</td>\n",
              "      <td>-16.050</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>150</td>\n",
              "      <td>-71.053</td>\n",
              "      <td>42.367</td>\n",
              "      <td>3.000</td>\n",
              "      <td>-75.928</td>\n",
              "      <td>-9.228</td>\n",
              "      <td>-44.980</td>\n",
              "      <td>14.000</td>\n",
              "      <td>0.000</td>\n",
              "      <td>-90.310</td>\n",
              "      <td>...</td>\n",
              "      <td>2.214</td>\n",
              "      <td>-80.420</td>\n",
              "      <td>-14.273</td>\n",
              "      <td>-96.010</td>\n",
              "      <td>-10.510</td>\n",
              "      <td>-116.840</td>\n",
              "      <td>-16.050</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>200</td>\n",
              "      <td>-71.053</td>\n",
              "      <td>42.367</td>\n",
              "      <td>3.000</td>\n",
              "      <td>-75.340</td>\n",
              "      <td>-9.689</td>\n",
              "      <td>-45.404</td>\n",
              "      <td>11.429</td>\n",
              "      <td>0.000</td>\n",
              "      <td>-90.310</td>\n",
              "      <td>...</td>\n",
              "      <td>2.250</td>\n",
              "      <td>-81.380</td>\n",
              "      <td>-11.560</td>\n",
              "      <td>-96.010</td>\n",
              "      <td>-10.510</td>\n",
              "      <td>-116.840</td>\n",
              "      <td>-16.050</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>317855</th>\n",
              "      <td>15892750</td>\n",
              "      <td>-71.053</td>\n",
              "      <td>42.367</td>\n",
              "      <td>0.000</td>\n",
              "      <td>-69.242</td>\n",
              "      <td>-8.337</td>\n",
              "      <td>-39.939</td>\n",
              "      <td>7.381</td>\n",
              "      <td>8.571</td>\n",
              "      <td>-86.680</td>\n",
              "      <td>...</td>\n",
              "      <td>0.000</td>\n",
              "      <td>-80.758</td>\n",
              "      <td>-17.966</td>\n",
              "      <td>-99.500</td>\n",
              "      <td>-10.510</td>\n",
              "      <td>-109.700</td>\n",
              "      <td>-12.280</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>317856</th>\n",
              "      <td>15892800</td>\n",
              "      <td>-71.053</td>\n",
              "      <td>42.367</td>\n",
              "      <td>0.000</td>\n",
              "      <td>-69.242</td>\n",
              "      <td>-8.337</td>\n",
              "      <td>-39.939</td>\n",
              "      <td>7.381</td>\n",
              "      <td>8.571</td>\n",
              "      <td>-86.680</td>\n",
              "      <td>...</td>\n",
              "      <td>0.000</td>\n",
              "      <td>-80.758</td>\n",
              "      <td>-17.966</td>\n",
              "      <td>-99.500</td>\n",
              "      <td>-10.510</td>\n",
              "      <td>-109.700</td>\n",
              "      <td>-12.280</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>317857</th>\n",
              "      <td>15892850</td>\n",
              "      <td>-71.053</td>\n",
              "      <td>42.367</td>\n",
              "      <td>0.000</td>\n",
              "      <td>-69.242</td>\n",
              "      <td>-8.337</td>\n",
              "      <td>-39.939</td>\n",
              "      <td>7.381</td>\n",
              "      <td>8.571</td>\n",
              "      <td>-86.680</td>\n",
              "      <td>...</td>\n",
              "      <td>0.000</td>\n",
              "      <td>-80.758</td>\n",
              "      <td>-17.966</td>\n",
              "      <td>-99.500</td>\n",
              "      <td>-10.510</td>\n",
              "      <td>-109.700</td>\n",
              "      <td>-12.280</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>317858</th>\n",
              "      <td>15892900</td>\n",
              "      <td>-71.053</td>\n",
              "      <td>42.367</td>\n",
              "      <td>0.000</td>\n",
              "      <td>-68.890</td>\n",
              "      <td>-8.481</td>\n",
              "      <td>-39.912</td>\n",
              "      <td>10.048</td>\n",
              "      <td>5.571</td>\n",
              "      <td>-86.680</td>\n",
              "      <td>...</td>\n",
              "      <td>0.000</td>\n",
              "      <td>-80.750</td>\n",
              "      <td>-16.001</td>\n",
              "      <td>-99.500</td>\n",
              "      <td>-10.510</td>\n",
              "      <td>-109.700</td>\n",
              "      <td>-12.280</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>317859</th>\n",
              "      <td>15892950</td>\n",
              "      <td>-71.053</td>\n",
              "      <td>42.367</td>\n",
              "      <td>0.000</td>\n",
              "      <td>-68.890</td>\n",
              "      <td>-8.481</td>\n",
              "      <td>-39.912</td>\n",
              "      <td>10.048</td>\n",
              "      <td>5.571</td>\n",
              "      <td>-86.680</td>\n",
              "      <td>...</td>\n",
              "      <td>0.000</td>\n",
              "      <td>-80.750</td>\n",
              "      <td>-16.001</td>\n",
              "      <td>-99.500</td>\n",
              "      <td>-10.510</td>\n",
              "      <td>-109.700</td>\n",
              "      <td>-12.280</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "      <td>0.000</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>317860 rows × 24 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-4382df32-7f26-4bbe-a23f-5989289e9038')\"\n",
              "              title=\"Convert this dataframe to an interactive table.\"\n",
              "              style=\"display:none;\">\n",
              "        \n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "      \n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "      <script>\n",
              "        const buttonEl =\n",
              "          document.querySelector('#df-4382df32-7f26-4bbe-a23f-5989289e9038 button.colab-df-convert');\n",
              "        buttonEl.style.display =\n",
              "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "        async function convertToInteractive(key) {\n",
              "          const element = document.querySelector('#df-4382df32-7f26-4bbe-a23f-5989289e9038');\n",
              "          const dataTable =\n",
              "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                     [key], {});\n",
              "          if (!dataTable) return;\n",
              "\n",
              "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "            + ' to learn more about interactive tables.';\n",
              "          element.innerHTML = '';\n",
              "          dataTable['output_type'] = 'display_data';\n",
              "          await google.colab.output.renderOutput(dataTable, element);\n",
              "          const docLink = document.createElement('div');\n",
              "          docLink.innerHTML = docLinkHtml;\n",
              "          element.appendChild(docLink);\n",
              "        }\n",
              "      </script>\n",
              "    </div>\n",
              "  </div>\n",
              "  "
            ]
          },
          "metadata": {}
        }
      ],
      "source": [
        "class MyDataset:\n",
        "    def __init__(self, PARAMETERS:dict):\n",
        "        self.PARAMETERS = PARAMETERS\n",
        "        self.df = self.read_data()     \n",
        "        \n",
        "        if self.PARAMETERS[\"binary\"] == True:\n",
        "            self.convert_binary_label()\n",
        "            \n",
        "        self.lte_le = preprocessing.LabelEncoder()\n",
        "        self.nr_le = preprocessing.LabelEncoder()\n",
        "        self.all_le = preprocessing.LabelEncoder()\n",
        "        self.encode_label()\n",
        "            \n",
        "        if self.PARAMETERS[\"sample_rate\"] != 0:\n",
        "            self.sampled_df = self.sampling()\n",
        "    \n",
        "    \n",
        "    def read_data(self):\n",
        "        data_path = path.join(self.PARAMETERS[\"data_folder\"], self.PARAMETERS[\"file_name\"] + '.csv')\n",
        "        df = pd.read_csv(data_path, low_memory=False)\n",
        "        \n",
        "        # Drop the first several rows\n",
        "        df = df[3000:].reset_index(drop=True)                                   \n",
        "        \n",
        "        # Convert '2021-06-15 15:07:18.554116' to 1623769638554.116 (ms)\n",
        "        df['TIME_STAMP'] = pd.to_datetime(df['TIME_STAMP'])\n",
        "        df['OLD_TIME_STAMP'] = df['TIME_STAMP']\n",
        "        df[\"TIME_STAMP\"] = (df[\"TIME_STAMP\"] - pd.Timestamp(\"1970-01-01\")) // pd.Timedelta('1us')\n",
        "        df[\"TIME_STAMP\"] = (df[\"TIME_STAMP\"] - df[\"TIME_STAMP\"][0])/1000\n",
        "        \n",
        "        # Merge \"lte_ho_category\" and \"nr_ho_category\"\n",
        "        df[\"all_ho_category\"] = df[\"lte_ho_category\"] + \" \" + df[\"nr_ho_category\"]\n",
        "\n",
        "        return df\n",
        "    \n",
        "    \n",
        "    def convert_binary_label(self):\n",
        "        df = self.df\n",
        "        df.loc[~(self.df[\"lte_ho_category\"] == \"no_handover\"), \"lte_ho_category\"] = 1\n",
        "        df.loc[self.df[\"lte_ho_category\"] == \"no_handover\", \"lte_ho_category\"] = 0\n",
        "        df.loc[~(self.df[\"nr_ho_category\"] == \"no_handover\"), \"nr_ho_category\"] = 1\n",
        "        df.loc[self.df[\"nr_ho_category\"] == \"no_handover\", \"nr_ho_category\"] = 0\n",
        "        df.loc[~(self.df[\"all_ho_category\"] == \"no_handover no_handover\"), \"all_ho_category\"] = 1\n",
        "        df.loc[self.df[\"all_ho_category\"] == \"no_handover no_handover\", \"all_ho_category\"] = 0\n",
        "        self.df = df\n",
        "        \n",
        "        \n",
        "    def encode_label(self):\n",
        "        df = self.df\n",
        "        self.lte_le = preprocessing.LabelEncoder().fit(df[\"lte_ho_category\"])\n",
        "        df.loc[:, \"lte_ho_category\"] = self.lte_le.transform(df.loc[:, \"lte_ho_category\"])\n",
        "        self.nr_le = preprocessing.LabelEncoder().fit(df[\"nr_ho_category\"])\n",
        "        df.loc[:, \"nr_ho_category\"] = self.nr_le.transform(df.loc[:, \"nr_ho_category\"])\n",
        "        self.all_le = preprocessing.LabelEncoder().fit(df[\"all_ho_category\"])\n",
        "        df.loc[:, \"all_ho_category\"] = self.all_le.transform(df.loc[:, \"all_ho_category\"])\n",
        "        self.df = df\n",
        "\n",
        "\n",
        "    def sampling(self):\n",
        "        df = self.df\n",
        "        sample_rate = self.PARAMETERS[\"sample_rate\"]  # default 50 ms\n",
        "        bins = range(int(df['TIME_STAMP'].min()), int(df['TIME_STAMP'].max())+1, sample_rate)\n",
        "        df['r_time'] = pd.cut(df['TIME_STAMP'], bins=bins, labels=bins[:-1], include_lowest=True)\n",
        "        sampled_df = df.groupby('r_time').agg(lon=('lon', np.mean), \n",
        "                                              lat=('lat', np.mean),\n",
        "                                              speed=('speed', np.mean),\n",
        "                                              lte_serv_rsrp=('lte_serv_rsrp', np.mean),\n",
        "                                              lte_serv_rsrq=('lte_serv_rsrq', np.mean),\n",
        "                                              lte_serv_rssi=('lte_serv_rssi', np.mean),\n",
        "                                              lte_cqi0=('lte_cqi0', np.mean),\n",
        "                                              lte_cqi1=('lte_cqi1', np.mean),\n",
        "                                              nr_serv_rsrp=('nr_serv_rsrp', np.mean),\n",
        "                                              nr_serv_rsrq=('nr_serv_rsrq', np.mean),\n",
        "                                              nr_serv_sinr=('nr_serv_sinr', np.mean),\n",
        "                                              nr_cqi=('nr_cqi', np.mean),\n",
        "                                              dl_tput=('dl_tput', np.mean),\n",
        "                                              ul_tput=('ul_tput', np.mean),\n",
        "                                              lte_nbr_rsrp=('lte_nbr_rsrp', np.mean),\n",
        "                                              lte_nbr_rsrq=('lte_nbr_rsrq', np.mean),\n",
        "                                              irat_rsrp=('irat_rsrp', np.mean),\n",
        "                                              irat_rsrq=('irat_rsrq', np.mean),\n",
        "                                              nr_nbr_rsrp=('nr_nbr_rsrp', np.mean),\n",
        "                                              nr_nbr_rsrq=('nr_nbr_rsrq', np.mean),\n",
        "                                              lte_ho_category=('lte_ho_category', np.max),\n",
        "                                              nr_ho_category=('nr_ho_category', np.max),\n",
        "                                              all_ho_category=('all_ho_category', np.max),) \n",
        "        # Padding 4*sample_rate=200 ms\n",
        "        for col in sampled_df:\n",
        "            if col ==\"r_time\":\n",
        "                continue\n",
        "            else:\n",
        "                sampled_df[col].fillna(method='ffill', limit=4, inplace=True)\n",
        "                \n",
        "        sampled_df.reset_index(level=0, inplace=True)\n",
        "        return sampled_df\n",
        "\n",
        "\n",
        "    def get_df(self):\n",
        "        return self.df\n",
        "\n",
        "\n",
        "    def get_sampled_df(self):\n",
        "        return self.sampled_df\n",
        "\n",
        "\n",
        "####################################\n",
        "### Read and sample the database\n",
        "####################################\n",
        "Dataset = MyDataset(PARAMETERS)\n",
        "df = Dataset.get_df()\n",
        "print(df.columns)\n",
        "\n",
        "sampled_df = Dataset.get_sampled_df()\n",
        "display(sampled_df)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "id": "maritime-northwest",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "maritime-northwest",
        "outputId": "54524b28-eede-4146-b77c-934b984a3870"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\n",
            "Processing train set: \n",
            "\n",
            "seq_ho_arr shape:  (2260, 40)\n",
            "seq_neg_arr shape:  (7464, 40)\n",
            "data_seq_point_ind shape:  (9724, 40)\n",
            "Processing:  0 / 9724\n",
            "Processing:  500 / 9724\n",
            "Processing:  1000 / 9724\n",
            "Processing:  1500 / 9724\n",
            "Processing:  2000 / 9724\n",
            "Processing:  2500 / 9724\n",
            "Processing:  3000 / 9724\n",
            "Processing:  3500 / 9724\n",
            "Processing:  4000 / 9724\n",
            "Processing:  4500 / 9724\n",
            "Processing:  5000 / 9724\n",
            "Processing:  5500 / 9724\n",
            "Processing:  6000 / 9724\n",
            "Processing:  6500 / 9724\n",
            "Processing:  7000 / 9724\n",
            "Processing:  7500 / 9724\n",
            "Processing:  8000 / 9724\n",
            "Processing:  8500 / 9724\n",
            "Processing:  9000 / 9724\n",
            "Processing:  9500 / 9724\n",
            "\n",
            "X_data_seq shape:  (9724, 40, 17)  y_data_seq shape:  (9724, 40)\n",
            "\n",
            "Processing test set: \n",
            "\n",
            "seq_ho_arr shape:  (492, 40)\n",
            "seq_all_arr shape:  (10000, 40)\n",
            "data_seq_point_ind shape:  (10492, 40)\n",
            "Processing:  0 / 10492\n",
            "Processing:  500 / 10492\n",
            "Processing:  1000 / 10492\n",
            "Processing:  1500 / 10492\n",
            "Processing:  2000 / 10492\n",
            "Processing:  2500 / 10492\n",
            "Processing:  3000 / 10492\n",
            "Processing:  3500 / 10492\n",
            "Processing:  4000 / 10492\n",
            "Processing:  4500 / 10492\n",
            "Processing:  5000 / 10492\n",
            "Processing:  5500 / 10492\n",
            "Processing:  6000 / 10492\n",
            "Processing:  6500 / 10492\n",
            "Processing:  7000 / 10492\n",
            "Processing:  7500 / 10492\n",
            "Processing:  8000 / 10492\n",
            "Processing:  8500 / 10492\n",
            "Processing:  9000 / 10492\n",
            "Processing:  9500 / 10492\n",
            "Processing:  10000 / 10492\n",
            "\n",
            "X_data_seq shape:  (10492, 40, 17)  y_data_seq shape:  (10492, 40)\n",
            "\n",
            "X_train_seq shape:  (9724, 40, 17)  y_train_seq shape:  (9724, 40)\n",
            "X_test_seq shape:  (10492, 40, 17)  y_train_seq shape:  (10492, 40)\n"
          ]
        }
      ],
      "source": [
        "def sample_ho_seq(df, PARAMETERS):\n",
        "    ho_locs = np.where(df[PARAMETERS[\"label\"]] == 1)[0]\n",
        "    nan_locs = np.where(df[PARAMETERS[\"label\"]] != df[PARAMETERS[\"label\"]])[0]\n",
        "    \n",
        "    ### generate ho sequences from df\n",
        "    # Total 40 time slots in each sequence, each HO contributes 5 sequences \n",
        "    seq_ho_list = []\n",
        "    previous_num_points = np.random.randint(40, size=(len(ho_locs), 10))  \n",
        "    for i, ho_loc in enumerate(ho_locs):\n",
        "        for pre_num in previous_num_points[i]:\n",
        "            pre_loc = ho_loc-pre_num\n",
        "            beh_loc = pre_loc+40\n",
        "            seq_ho_list.append(range(pre_loc, beh_loc))\n",
        "\n",
        "    # check no Nan value in seq\n",
        "    seq_save = []\n",
        "    for i, seq in enumerate(seq_ho_list):\n",
        "        seq_min = seq[0]\n",
        "        seq_max = seq[-1]\n",
        "        if len(np.where((nan_locs>=seq_min) & (nan_locs<=seq_max))[0])==0 and \\\n",
        "          seq_max<len(df):\n",
        "            seq_save.append(i)\n",
        "\n",
        "    seq_ho_arr = np.array(seq_ho_list)[seq_save]\n",
        "    print(\"\\nseq_ho_arr shape: \", seq_ho_arr.shape)\n",
        "    return seq_ho_arr\n",
        "\n",
        "\n",
        "def sample_neg_seq(df, PARAMETERS):\n",
        "    ho_locs = np.where(df[PARAMETERS[\"label\"]] == 1)[0]\n",
        "    nan_locs = np.where(df[PARAMETERS[\"label\"]] != df[PARAMETERS[\"label\"]])[0]\n",
        "\n",
        "    ### generate neg sequences from df\n",
        "    seq_neg_list = []\n",
        "    rand_arr = np.random.randint(len(df), size=PARAMETERS[\"neg_size\"]) \n",
        "    for pre_loc in rand_arr:\n",
        "        beh_loc = pre_loc+40                                                                                 \n",
        "        seq_neg_list.append(range(pre_loc, beh_loc))\n",
        "\n",
        "    # check no Nan and ho in seq\n",
        "    seq_save = []\n",
        "    for i, seq in enumerate(seq_neg_list):\n",
        "        seq_min = seq[0]\n",
        "        seq_max = seq[-1]\n",
        "        if len(np.where((nan_locs>=seq_min) & (nan_locs<=seq_max))[0])==0 and \\\n",
        "          len(np.where((ho_locs>=seq_min) & (ho_locs<=seq_max))[0])==0 and \\\n",
        "          seq_max<len(df):\n",
        "                seq_save.append(i)    \n",
        "\n",
        "    seq_neg_arr = np.array(seq_neg_list)[seq_save]\n",
        "    print(\"seq_neg_arr shape: \", seq_neg_arr.shape)\n",
        "    return seq_neg_arr\n",
        "\n",
        "\n",
        "def sample_seq(df, PARAMETERS):\n",
        "    # Input: df should start from index 0\n",
        "\n",
        "    ### generate data sequences \n",
        "    seq_ho_arr = sample_ho_seq(df, PARAMETERS)\n",
        "    seq_neg_arr = sample_neg_seq(df, PARAMETERS)\n",
        "    data_seq_point_ind = np.concatenate((seq_ho_arr, seq_neg_arr), axis=0)\n",
        "    print(\"data_seq_point_ind shape: \", data_seq_point_ind.shape)\n",
        "\n",
        "    ### retrieve the features\n",
        "    data_seq = []\n",
        "    for i, seq in enumerate(data_seq_point_ind):\n",
        "        if i%500==0:\n",
        "            print(\"Processing: \", i,\"/\",len(data_seq_point_ind))\n",
        "        data_seq.append(np.array(df.loc[seq, PARAMETERS[\"feature_list\"]+[PARAMETERS[\"label\"]]]))\n",
        "    X_data_seq, y_data_seq = np.split(np.array(data_seq), [len(PARAMETERS[\"feature_list\"])], axis=2)\n",
        "    y_data_seq = y_data_seq.squeeze()\n",
        "    print(\"\\nX_data_seq shape: \", X_data_seq.shape, \" y_data_seq shape: \", y_data_seq.shape)\n",
        "\n",
        "    ### split the dataset into train and test\n",
        "    X_data_seq_ind = np.arange(len(X_data_seq))\n",
        "    np.random.shuffle(X_data_seq_ind)\n",
        "\n",
        "    if PARAMETERS[\"split_mode\"] == \"all_neg_sampling\":\n",
        "        num_train_samples = int(PARAMETERS[\"train_ratio\"]*len(X_data_seq_ind))\n",
        "    elif PARAMETERS[\"split_mode\"] == \"train_neg_sampling\":\n",
        "        num_train_samples = int(len(X_data_seq_ind))\n",
        "    X_train_seq = X_data_seq[X_data_seq_ind][:num_train_samples]\n",
        "    y_train_seq = y_data_seq[X_data_seq_ind][:num_train_samples]\n",
        "    X_test_seq = X_data_seq[X_data_seq_ind][num_train_samples:]\n",
        "    y_test_seq = y_data_seq[X_data_seq_ind][num_train_samples:]\n",
        "    return X_train_seq, y_train_seq, X_test_seq, y_test_seq\n",
        "\n",
        "\n",
        "def convert_df_to_seq(df, PARAMETERS):\n",
        "    df = df.reset_index(drop=True)\n",
        "    nan_loc = np.where(df[PARAMETERS[\"label\"]] != df[PARAMETERS[\"label\"]])[0]\n",
        "\n",
        "    ### generate ind sequences \n",
        "    seq_ho_arr = sample_ho_seq(df, PARAMETERS)\n",
        "\n",
        "    seq_all_list = []\n",
        "    for ind in range(len(df)-40):                                                                             \n",
        "        if len(np.where((nan_loc>=ind) & (nan_loc<=ind+40))[0])==0:\n",
        "            seq_all_list.append(range(ind, ind+40))\n",
        "    random_list = random.sample(range(len(seq_all_list)), 10000)\n",
        "    seq_all_arr = np.array(seq_all_list)[random_list]\n",
        "    print(\"seq_all_arr shape: \", seq_all_arr.shape)\n",
        "\n",
        "    data_seq_point_ind = np.concatenate((seq_ho_arr, seq_all_arr), axis=0)\n",
        "    print(\"data_seq_point_ind shape: \", data_seq_point_ind.shape)\n",
        "\n",
        "\n",
        "    ### retrieve the features\n",
        "    data_seq = []\n",
        "    for i, seq in enumerate(data_seq_point_ind):\n",
        "        if i%500==0:\n",
        "            print(\"Processing: \", i,\"/\",len(data_seq_point_ind))\n",
        "        data_seq.append(np.array(df.loc[seq, PARAMETERS[\"feature_list\"]+[PARAMETERS[\"label\"]]]))\n",
        "    X_data_seq, y_data_seq = np.split(np.array(data_seq), [len(PARAMETERS[\"feature_list\"])], axis=2)\n",
        "    y_data_seq = y_data_seq.squeeze()\n",
        "    print(\"\\nX_data_seq shape: \", X_data_seq.shape, \" y_data_seq shape: \", y_data_seq.shape)\n",
        "\n",
        "    return X_data_seq, y_data_seq\n",
        "\n",
        "\n",
        "####################################\n",
        "### Generate data sequences for train and test \n",
        "####################################\n",
        "if PARAMETERS[\"split_mode\"] == \"all_neg_sampling\":\n",
        "    # Do the negative sampling on the whole dataset\n",
        "    # Split the sampled sequences for train and test\n",
        "    X_train_seq, y_train_seq, X_test_seq, y_test_seq = sample_seq(sampled_df, PARAMETERS)\n",
        "\n",
        "\n",
        "if PARAMETERS[\"split_mode\"] == \"train_neg_sampling\":\n",
        "    # Split the dataset into training and test part\n",
        "    # Do the negative sampling on the traing dataset, \n",
        "    # Use the sampled sequences for training, and the whole test dataset for test \n",
        "    num_train_samples = int(PARAMETERS[\"train_ratio\"]*len(sampled_df))\n",
        "    train_df = sampled_df[:num_train_samples]\n",
        "    test_df = sampled_df[num_train_samples:]\n",
        "    print(\"\\nProcessing train set: \")\n",
        "    X_train_seq, y_train_seq, _, _ = sample_seq(train_df, PARAMETERS)\n",
        "    print(\"\\nProcessing test set: \")\n",
        "    X_test_seq, y_test_seq = convert_df_to_seq(test_df, PARAMETERS)\n",
        "\n",
        "\n",
        "print(\"\\nX_train_seq shape: \", X_train_seq.shape, \" y_train_seq shape: \", y_train_seq.shape)\n",
        "print(\"X_test_seq shape: \", X_test_seq.shape, \" y_train_seq shape: \", y_test_seq.shape)"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Train and evaluate the models\n",
        "* **GradientBoostingClassifier**\n",
        "\n",
        "The following models can also be tested:\n",
        "* DecisionTreeClassifier\n",
        "* RandomForestClassifier\n",
        "* AdaBoostClassifier"
      ],
      "metadata": {
        "id": "UQbacTsmFkFm"
      },
      "id": "UQbacTsmFkFm"
    },
    {
      "cell_type": "code",
      "source": [
        "# further split the traing data for model \n",
        "X_train = np.hsplit(X_train_seq, 2)[0].reshape(len(X_train_seq), -1)  # concatenate\n",
        "y_train = np.hsplit(y_train_seq, 2)[1].max(axis=1)\n",
        "print(\"X_train shape: \", X_train.shape, \" y_train shape: \", y_train.shape)\n",
        "\n",
        "# further split the test data for model \n",
        "X_test  = np.hsplit(X_test_seq, 2)[0].reshape(len(X_test_seq), -1)    # concatenate\n",
        "y_test = np.hsplit(y_test_seq, 2)[1].max(axis=1)\n",
        "print(\"X_test shape: \", X_test.shape, \" y_test shape: \", y_test.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "kuXc63vTeDYP",
        "outputId": "94e38d6a-0e0c-40ec-ec92-84fae0e4569b"
      },
      "id": "kuXc63vTeDYP",
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "X_train shape:  (9724, 340)  y_train shape:  (9724,)\n",
            "X_test shape:  (10492, 340)  y_test shape:  (10492,)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "##########################\n",
        "### Init the model\n",
        "##########################\n",
        "# model = tree.DecisionTreeClassifier()\n",
        "# model = RandomForestClassifier()\n",
        "# model = AdaBoostClassifier()\n",
        "model = GradientBoostingClassifier()\n",
        "\n",
        "\n",
        "##########################\n",
        "### Fit \n",
        "##########################\n",
        "model.fit(X=X_train, y=y_train)  \n",
        "\n",
        "\n",
        "##########################\n",
        "### Evaluation\n",
        "##########################\n",
        "y_pred = model.predict(X_test)\n",
        "y_true = y_test\n",
        "\n",
        "bi_f1 =  f1_score(y_true, y_pred)  # , average='macro'\n",
        "bi_recall = recall_score(y_true, y_pred)\n",
        "bi_precision = precision_score(y_true, y_pred)\n",
        "bi_acc = accuracy_score(y_true, y_pred)\n",
        "print(\"Binary classification: F1-score: {:.3} Recall: {:.3} Precision: {:.3} Accuracy: {:.3}\".format(bi_f1, bi_recall, bi_precision, bi_acc))\n",
        "\n",
        "# Plot the confusion matrix\n",
        "plt.figure(figsize=(8, 8))\n",
        "cfm = confusion_matrix(y_true, y_pred)  # labels=np.unique(lte_le.inverse_transform(y_true)\n",
        "ax = sn.heatmap(cfm, annot=True, annot_kws={\"size\": 12}, fmt='.20g')\n",
        "plt.xlabel(\"predicted label\")\n",
        "plt.ylabel(\"true label\")\n",
        "plt.show()\n",
        "\n",
        "# Plot ROC curve\n",
        "y_probas = model.predict_proba(X_test)\n",
        "plot_roc_curve(model, X_test, y_test) \n",
        "plt.show()"
      ],
      "metadata": {
        "id": "E27go599FkLq",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 831
        },
        "outputId": "313ccd26-e4bf-41b3-8d28-b9e57a644f95"
      },
      "id": "E27go599FkLq",
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Binary classification: F1-score: 0.52 Recall: 0.467 Precision: 0.586 Accuracy: 0.957\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 576x576 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/utils/deprecation.py:87: FutureWarning: Function plot_roc_curve is deprecated; Function :func:`plot_roc_curve` is deprecated in 1.0 and will be removed in 1.2. Use one of the class methods: :meth:`sklearn.metric.RocCurveDisplay.from_predictions` or :meth:`sklearn.metric.RocCurveDisplay.from_estimator`.\n",
            "  warnings.warn(msg, category=FutureWarning)\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    }
  ],
  "metadata": {
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    },
    "colab": {
      "name": "GBC.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "gpuClass": "standard"
  },
  "nbformat": 4,
  "nbformat_minor": 5
}