SVM - Results of hyperparameter search.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.svm import SVC\n",
    "from sklearn.metrics import accuracy_score, f1_score, confusion_matrix\n",
    "import joblib"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Hyper param search results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Gradient Boosting"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "dd = pd.read_csv(\"Hyperparam_search/SVM_GS_results.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <th>mean_fit_time</th>\n",
       "      <th>std_fit_time</th>\n",
       "      <th>mean_score_time</th>\n",
       "      <th>std_score_time</th>\n",
       "      <th>param_C</th>\n",
       "      <th>param_gamma</th>\n",
       "      <th>param_kernel</th>\n",
       "      <th>params</th>\n",
       "      <th>split0_test_accuracy</th>\n",
       "      <th>...</th>\n",
       "      <th>std_test_accuracy</th>\n",
       "      <th>rank_test_accuracy</th>\n",
       "      <th>split0_test_F1</th>\n",
       "      <th>split1_test_F1</th>\n",
       "      <th>split2_test_F1</th>\n",
       "      <th>split3_test_F1</th>\n",
       "      <th>split4_test_F1</th>\n",
       "      <th>mean_test_F1</th>\n",
       "      <th>std_test_F1</th>\n",
       "      <th>rank_test_F1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>7</td>\n",
       "      <td>126.222607</td>\n",
       "      <td>1.236996</td>\n",
       "      <td>26.665054</td>\n",
       "      <td>0.145072</td>\n",
       "      <td>10</td>\n",
       "      <td>scale</td>\n",
       "      <td>rbf</td>\n",
       "      <td>{'C': 10, 'gamma': 'scale', 'kernel': 'rbf'}</td>\n",
       "      <td>0.834354</td>\n",
       "      <td>...</td>\n",
       "      <td>0.003933</td>\n",
       "      <td>1</td>\n",
       "      <td>0.786809</td>\n",
       "      <td>0.795040</td>\n",
       "      <td>0.786164</td>\n",
       "      <td>0.797157</td>\n",
       "      <td>0.776030</td>\n",
       "      <td>0.788240</td>\n",
       "      <td>0.007499</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>91.794349</td>\n",
       "      <td>1.023251</td>\n",
       "      <td>27.488348</td>\n",
       "      <td>0.026995</td>\n",
       "      <td>1</td>\n",
       "      <td>scale</td>\n",
       "      <td>rbf</td>\n",
       "      <td>{'C': 1, 'gamma': 'scale', 'kernel': 'rbf'}</td>\n",
       "      <td>0.831068</td>\n",
       "      <td>...</td>\n",
       "      <td>0.003721</td>\n",
       "      <td>5</td>\n",
       "      <td>0.785199</td>\n",
       "      <td>0.792571</td>\n",
       "      <td>0.787630</td>\n",
       "      <td>0.795630</td>\n",
       "      <td>0.778996</td>\n",
       "      <td>0.788005</td>\n",
       "      <td>0.005799</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>100.721196</td>\n",
       "      <td>1.827760</td>\n",
       "      <td>28.011370</td>\n",
       "      <td>0.149500</td>\n",
       "      <td>1</td>\n",
       "      <td>0.1</td>\n",
       "      <td>rbf</td>\n",
       "      <td>{'C': 1, 'gamma': 0.1, 'kernel': 'rbf'}</td>\n",
       "      <td>0.833981</td>\n",
       "      <td>...</td>\n",
       "      <td>0.003891</td>\n",
       "      <td>3</td>\n",
       "      <td>0.786222</td>\n",
       "      <td>0.793233</td>\n",
       "      <td>0.784677</td>\n",
       "      <td>0.790927</td>\n",
       "      <td>0.773849</td>\n",
       "      <td>0.785782</td>\n",
       "      <td>0.006720</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>146.870323</td>\n",
       "      <td>2.050380</td>\n",
       "      <td>26.789975</td>\n",
       "      <td>0.469019</td>\n",
       "      <td>10</td>\n",
       "      <td>0.1</td>\n",
       "      <td>rbf</td>\n",
       "      <td>{'C': 10, 'gamma': 0.1, 'kernel': 'rbf'}</td>\n",
       "      <td>0.834354</td>\n",
       "      <td>...</td>\n",
       "      <td>0.003303</td>\n",
       "      <td>2</td>\n",
       "      <td>0.784750</td>\n",
       "      <td>0.785752</td>\n",
       "      <td>0.781722</td>\n",
       "      <td>0.789547</td>\n",
       "      <td>0.773095</td>\n",
       "      <td>0.782974</td>\n",
       "      <td>0.005536</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>11</td>\n",
       "      <td>178.989075</td>\n",
       "      <td>2.780193</td>\n",
       "      <td>25.920260</td>\n",
       "      <td>0.437622</td>\n",
       "      <td>25</td>\n",
       "      <td>scale</td>\n",
       "      <td>rbf</td>\n",
       "      <td>{'C': 25, 'gamma': 'scale', 'kernel': 'rbf'}</td>\n",
       "      <td>0.832188</td>\n",
       "      <td>...</td>\n",
       "      <td>0.003310</td>\n",
       "      <td>4</td>\n",
       "      <td>0.781156</td>\n",
       "      <td>0.783828</td>\n",
       "      <td>0.782027</td>\n",
       "      <td>0.788809</td>\n",
       "      <td>0.772894</td>\n",
       "      <td>0.781743</td>\n",
       "      <td>0.005157</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 25 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    Unnamed: 0  mean_fit_time  std_fit_time  mean_score_time  std_score_time  \\\n",
       "7            7     126.222607      1.236996        26.665054        0.145072   \n",
       "3            3      91.794349      1.023251        27.488348        0.026995   \n",
       "0            0     100.721196      1.827760        28.011370        0.149500   \n",
       "4            4     146.870323      2.050380        26.789975        0.469019   \n",
       "11          11     178.989075      2.780193        25.920260        0.437622   \n",
       "\n",
       "    param_C param_gamma param_kernel  \\\n",
       "7        10       scale          rbf   \n",
       "3         1       scale          rbf   \n",
       "0         1         0.1          rbf   \n",
       "4        10         0.1          rbf   \n",
       "11       25       scale          rbf   \n",
       "\n",
       "                                          params  split0_test_accuracy  ...  \\\n",
       "7   {'C': 10, 'gamma': 'scale', 'kernel': 'rbf'}              0.834354  ...   \n",
       "3    {'C': 1, 'gamma': 'scale', 'kernel': 'rbf'}              0.831068  ...   \n",
       "0        {'C': 1, 'gamma': 0.1, 'kernel': 'rbf'}              0.833981  ...   \n",
       "4       {'C': 10, 'gamma': 0.1, 'kernel': 'rbf'}              0.834354  ...   \n",
       "11  {'C': 25, 'gamma': 'scale', 'kernel': 'rbf'}              0.832188  ...   \n",
       "\n",
       "    std_test_accuracy  rank_test_accuracy  split0_test_F1  split1_test_F1  \\\n",
       "7            0.003933                   1        0.786809        0.795040   \n",
       "3            0.003721                   5        0.785199        0.792571   \n",
       "0            0.003891                   3        0.786222        0.793233   \n",
       "4            0.003303                   2        0.784750        0.785752   \n",
       "11           0.003310                   4        0.781156        0.783828   \n",
       "\n",
       "    split2_test_F1  split3_test_F1  split4_test_F1  mean_test_F1  std_test_F1  \\\n",
       "7         0.786164        0.797157        0.776030      0.788240     0.007499   \n",
       "3         0.787630        0.795630        0.778996      0.788005     0.005799   \n",
       "0         0.784677        0.790927        0.773849      0.785782     0.006720   \n",
       "4         0.781722        0.789547        0.773095      0.782974     0.005536   \n",
       "11        0.782027        0.788809        0.772894      0.781743     0.005157   \n",
       "\n",
       "    rank_test_F1  \n",
       "7              1  \n",
       "3              2  \n",
       "0              3  \n",
       "4              4  \n",
       "11             5  \n",
       "\n",
       "[5 rows x 25 columns]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dd.sort_values(by=[\"mean_test_F1\"], ascending=False).head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([  1,  10,  25,  50,  75, 100])"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.unique(dd[\"param_C\"].values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['0.001', '0.01', '0.1', 'scale'], dtype=object)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.unique(dd[\"param_gamma\"].values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "pgamma = np.array([0.001, 0.01, 0.07, 0.1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [],
   "source": [
    "C, G = np.meshgrid(np.unique(dd[\"param_C\"].values), pgamma)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[  1,  10,  25,  50,  75, 100],\n",
       "       [  1,  10,  25,  50,  75, 100],\n",
       "       [  1,  10,  25,  50,  75, 100],\n",
       "       [  1,  10,  25,  50,  75, 100]])"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "C"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.001, 0.001, 0.001, 0.001, 0.001, 0.001],\n",
       "       [0.01 , 0.01 , 0.01 , 0.01 , 0.01 , 0.01 ],\n",
       "       [0.07 , 0.07 , 0.07 , 0.07 , 0.07 , 0.07 ],\n",
       "       [0.1  , 0.1  , 0.1  , 0.1  , 0.1  , 0.1  ]])"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "G"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [],
   "source": [
    "grid = lambda df,lr,ne: df[df[\"param_C\"]==lr][df[\"param_gamma\"]==ne][\"mean_test_F1\"].values[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/Ghosty/.local/lib/python3.7/site-packages/ipykernel_launcher.py:1: UserWarning: Boolean Series key will be reindexed to match DataFrame index.\n",
      "  \"\"\"Entry point for launching an IPython kernel.\n"
     ]
    }
   ],
   "source": [
    "Y_1 = [grid(dd,n,'0.001') for n in np.unique(dd[\"param_C\"].values)]\n",
    "Y_3 = [grid(dd,n,\"scale\") for n in np.unique(dd[\"param_C\"].values)]\n",
    "Y_2 = [grid(dd,n,'0.01') for n in np.unique(dd[\"param_C\"].values)]\n",
    "Y_4 = [grid(dd,n,'0.1') for n in np.unique(dd[\"param_C\"].values)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [],
   "source": [
    "Z = np.vstack((Y_1, Y_2,Y_3,Y_4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10,7))\n",
    "plt.contourf(C, G, Z, 20, cmap='RdGy')\n",
    "plt.colorbar()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "dc = pd.read_csv(\"Hyperparam_search/SVM_GS_CVAE_elu_results.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <th>mean_fit_time</th>\n",
       "      <th>std_fit_time</th>\n",
       "      <th>mean_score_time</th>\n",
       "      <th>std_score_time</th>\n",
       "      <th>param_C</th>\n",
       "      <th>param_gamma</th>\n",
       "      <th>param_kernel</th>\n",
       "      <th>params</th>\n",
       "      <th>split0_test_accuracy</th>\n",
       "      <th>...</th>\n",
       "      <th>std_test_accuracy</th>\n",
       "      <th>rank_test_accuracy</th>\n",
       "      <th>split0_test_F1</th>\n",
       "      <th>split1_test_F1</th>\n",
       "      <th>split2_test_F1</th>\n",
       "      <th>split3_test_F1</th>\n",
       "      <th>split4_test_F1</th>\n",
       "      <th>mean_test_F1</th>\n",
       "      <th>std_test_F1</th>\n",
       "      <th>rank_test_F1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>97.680438</td>\n",
       "      <td>5.791368</td>\n",
       "      <td>28.705295</td>\n",
       "      <td>2.223985</td>\n",
       "      <td>1</td>\n",
       "      <td>0.1</td>\n",
       "      <td>rbf</td>\n",
       "      <td>{'C': 1, 'gamma': 0.1, 'kernel': 'rbf'}</td>\n",
       "      <td>0.838237</td>\n",
       "      <td>...</td>\n",
       "      <td>0.004028</td>\n",
       "      <td>3</td>\n",
       "      <td>0.789364</td>\n",
       "      <td>0.802397</td>\n",
       "      <td>0.791160</td>\n",
       "      <td>0.802246</td>\n",
       "      <td>0.783477</td>\n",
       "      <td>0.793729</td>\n",
       "      <td>0.007462</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>85.976293</td>\n",
       "      <td>3.115731</td>\n",
       "      <td>26.638293</td>\n",
       "      <td>0.225740</td>\n",
       "      <td>1</td>\n",
       "      <td>scale</td>\n",
       "      <td>rbf</td>\n",
       "      <td>{'C': 1, 'gamma': 'scale', 'kernel': 'rbf'}</td>\n",
       "      <td>0.835474</td>\n",
       "      <td>...</td>\n",
       "      <td>0.003538</td>\n",
       "      <td>5</td>\n",
       "      <td>0.787857</td>\n",
       "      <td>0.800408</td>\n",
       "      <td>0.794434</td>\n",
       "      <td>0.799481</td>\n",
       "      <td>0.782970</td>\n",
       "      <td>0.793031</td>\n",
       "      <td>0.006723</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>7</td>\n",
       "      <td>128.012036</td>\n",
       "      <td>6.040978</td>\n",
       "      <td>26.609327</td>\n",
       "      <td>1.146093</td>\n",
       "      <td>10</td>\n",
       "      <td>scale</td>\n",
       "      <td>rbf</td>\n",
       "      <td>{'C': 10, 'gamma': 'scale', 'kernel': 'rbf'}</td>\n",
       "      <td>0.840179</td>\n",
       "      <td>...</td>\n",
       "      <td>0.003922</td>\n",
       "      <td>1</td>\n",
       "      <td>0.794062</td>\n",
       "      <td>0.797575</td>\n",
       "      <td>0.794132</td>\n",
       "      <td>0.794100</td>\n",
       "      <td>0.779205</td>\n",
       "      <td>0.791815</td>\n",
       "      <td>0.006447</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>159.802753</td>\n",
       "      <td>8.510879</td>\n",
       "      <td>26.762137</td>\n",
       "      <td>0.839174</td>\n",
       "      <td>10</td>\n",
       "      <td>0.1</td>\n",
       "      <td>rbf</td>\n",
       "      <td>{'C': 10, 'gamma': 0.1, 'kernel': 'rbf'}</td>\n",
       "      <td>0.838536</td>\n",
       "      <td>...</td>\n",
       "      <td>0.004212</td>\n",
       "      <td>2</td>\n",
       "      <td>0.786326</td>\n",
       "      <td>0.795740</td>\n",
       "      <td>0.789696</td>\n",
       "      <td>0.793142</td>\n",
       "      <td>0.778711</td>\n",
       "      <td>0.788724</td>\n",
       "      <td>0.005928</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>11</td>\n",
       "      <td>170.444028</td>\n",
       "      <td>4.143445</td>\n",
       "      <td>25.406041</td>\n",
       "      <td>0.597963</td>\n",
       "      <td>25</td>\n",
       "      <td>scale</td>\n",
       "      <td>rbf</td>\n",
       "      <td>{'C': 25, 'gamma': 'scale', 'kernel': 'rbf'}</td>\n",
       "      <td>0.836968</td>\n",
       "      <td>...</td>\n",
       "      <td>0.004475</td>\n",
       "      <td>6</td>\n",
       "      <td>0.784907</td>\n",
       "      <td>0.792570</td>\n",
       "      <td>0.785336</td>\n",
       "      <td>0.789972</td>\n",
       "      <td>0.773142</td>\n",
       "      <td>0.785186</td>\n",
       "      <td>0.006672</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 25 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    Unnamed: 0  mean_fit_time  std_fit_time  mean_score_time  std_score_time  \\\n",
       "0            0      97.680438      5.791368        28.705295        2.223985   \n",
       "3            3      85.976293      3.115731        26.638293        0.225740   \n",
       "7            7     128.012036      6.040978        26.609327        1.146093   \n",
       "4            4     159.802753      8.510879        26.762137        0.839174   \n",
       "11          11     170.444028      4.143445        25.406041        0.597963   \n",
       "\n",
       "    param_C param_gamma param_kernel  \\\n",
       "0         1         0.1          rbf   \n",
       "3         1       scale          rbf   \n",
       "7        10       scale          rbf   \n",
       "4        10         0.1          rbf   \n",
       "11       25       scale          rbf   \n",
       "\n",
       "                                          params  split0_test_accuracy  ...  \\\n",
       "0        {'C': 1, 'gamma': 0.1, 'kernel': 'rbf'}              0.838237  ...   \n",
       "3    {'C': 1, 'gamma': 'scale', 'kernel': 'rbf'}              0.835474  ...   \n",
       "7   {'C': 10, 'gamma': 'scale', 'kernel': 'rbf'}              0.840179  ...   \n",
       "4       {'C': 10, 'gamma': 0.1, 'kernel': 'rbf'}              0.838536  ...   \n",
       "11  {'C': 25, 'gamma': 'scale', 'kernel': 'rbf'}              0.836968  ...   \n",
       "\n",
       "    std_test_accuracy  rank_test_accuracy  split0_test_F1  split1_test_F1  \\\n",
       "0            0.004028                   3        0.789364        0.802397   \n",
       "3            0.003538                   5        0.787857        0.800408   \n",
       "7            0.003922                   1        0.794062        0.797575   \n",
       "4            0.004212                   2        0.786326        0.795740   \n",
       "11           0.004475                   6        0.784907        0.792570   \n",
       "\n",
       "    split2_test_F1  split3_test_F1  split4_test_F1  mean_test_F1  std_test_F1  \\\n",
       "0         0.791160        0.802246        0.783477      0.793729     0.007462   \n",
       "3         0.794434        0.799481        0.782970      0.793031     0.006723   \n",
       "7         0.794132        0.794100        0.779205      0.791815     0.006447   \n",
       "4         0.789696        0.793142        0.778711      0.788724     0.005928   \n",
       "11        0.785336        0.789972        0.773142      0.785186     0.006672   \n",
       "\n",
       "    rank_test_F1  \n",
       "0              1  \n",
       "3              2  \n",
       "7              3  \n",
       "4              4  \n",
       "11             5  \n",
       "\n",
       "[5 rows x 25 columns]"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dc.sort_values(by=[\"mean_test_F1\"], ascending=False).head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/Ghosty/.local/lib/python3.7/site-packages/ipykernel_launcher.py:1: UserWarning: Boolean Series key will be reindexed to match DataFrame index.\n",
      "  \"\"\"Entry point for launching an IPython kernel.\n"
     ]
    }
   ],
   "source": [
    "Y_1 = [grid(dc,n,'0.001') for n in np.unique(dd[\"param_C\"].values)]\n",
    "Y_3 = [grid(dc,n,\"scale\") for n in np.unique(dd[\"param_C\"].values)]\n",
    "Y_2 = [grid(dc,n,'0.01') for n in np.unique(dd[\"param_C\"].values)]\n",
    "Y_4 = [grid(dc,n,'0.1') for n in np.unique(dd[\"param_C\"].values)]\n",
    "Z1 = np.vstack((Y_1, Y_2, Y_3, Y_4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10,7))\n",
    "plt.contourf(C, G, Z1, 20, cmap='RdGy')\n",
    "plt.colorbar()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "dr = pd.read_csv(\"Hyperparam_search/SVM_GS_LSTM_results.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <th>mean_fit_time</th>\n",
       "      <th>std_fit_time</th>\n",
       "      <th>mean_score_time</th>\n",
       "      <th>std_score_time</th>\n",
       "      <th>param_C</th>\n",
       "      <th>param_gamma</th>\n",
       "      <th>param_kernel</th>\n",
       "      <th>params</th>\n",
       "      <th>split0_test_accuracy</th>\n",
       "      <th>...</th>\n",
       "      <th>std_test_accuracy</th>\n",
       "      <th>rank_test_accuracy</th>\n",
       "      <th>split0_test_F1</th>\n",
       "      <th>split1_test_F1</th>\n",
       "      <th>split2_test_F1</th>\n",
       "      <th>split3_test_F1</th>\n",
       "      <th>split4_test_F1</th>\n",
       "      <th>mean_test_F1</th>\n",
       "      <th>std_test_F1</th>\n",
       "      <th>rank_test_F1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>23</td>\n",
       "      <td>136.975841</td>\n",
       "      <td>2.200803</td>\n",
       "      <td>22.996254</td>\n",
       "      <td>0.472610</td>\n",
       "      <td>100</td>\n",
       "      <td>scale</td>\n",
       "      <td>rbf</td>\n",
       "      <td>{'C': 100, 'gamma': 'scale', 'kernel': 'rbf'}</td>\n",
       "      <td>0.830246</td>\n",
       "      <td>...</td>\n",
       "      <td>0.003774</td>\n",
       "      <td>1</td>\n",
       "      <td>0.788064</td>\n",
       "      <td>0.799798</td>\n",
       "      <td>0.791047</td>\n",
       "      <td>0.793089</td>\n",
       "      <td>0.775469</td>\n",
       "      <td>0.789494</td>\n",
       "      <td>0.008003</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>19</td>\n",
       "      <td>120.700509</td>\n",
       "      <td>1.970945</td>\n",
       "      <td>22.701440</td>\n",
       "      <td>0.465616</td>\n",
       "      <td>75</td>\n",
       "      <td>scale</td>\n",
       "      <td>rbf</td>\n",
       "      <td>{'C': 75, 'gamma': 'scale', 'kernel': 'rbf'}</td>\n",
       "      <td>0.829500</td>\n",
       "      <td>...</td>\n",
       "      <td>0.003748</td>\n",
       "      <td>2</td>\n",
       "      <td>0.786916</td>\n",
       "      <td>0.798719</td>\n",
       "      <td>0.788953</td>\n",
       "      <td>0.792964</td>\n",
       "      <td>0.775946</td>\n",
       "      <td>0.788700</td>\n",
       "      <td>0.007543</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>15</td>\n",
       "      <td>104.998210</td>\n",
       "      <td>1.040820</td>\n",
       "      <td>23.312394</td>\n",
       "      <td>0.183782</td>\n",
       "      <td>50</td>\n",
       "      <td>scale</td>\n",
       "      <td>rbf</td>\n",
       "      <td>{'C': 50, 'gamma': 'scale', 'kernel': 'rbf'}</td>\n",
       "      <td>0.828305</td>\n",
       "      <td>...</td>\n",
       "      <td>0.003650</td>\n",
       "      <td>3</td>\n",
       "      <td>0.785875</td>\n",
       "      <td>0.797594</td>\n",
       "      <td>0.788175</td>\n",
       "      <td>0.793475</td>\n",
       "      <td>0.775783</td>\n",
       "      <td>0.788181</td>\n",
       "      <td>0.007425</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>11</td>\n",
       "      <td>88.963005</td>\n",
       "      <td>1.429340</td>\n",
       "      <td>23.226407</td>\n",
       "      <td>0.503023</td>\n",
       "      <td>25</td>\n",
       "      <td>scale</td>\n",
       "      <td>rbf</td>\n",
       "      <td>{'C': 25, 'gamma': 'scale', 'kernel': 'rbf'}</td>\n",
       "      <td>0.827035</td>\n",
       "      <td>...</td>\n",
       "      <td>0.003642</td>\n",
       "      <td>4</td>\n",
       "      <td>0.785165</td>\n",
       "      <td>0.797199</td>\n",
       "      <td>0.786614</td>\n",
       "      <td>0.794482</td>\n",
       "      <td>0.774684</td>\n",
       "      <td>0.787629</td>\n",
       "      <td>0.007915</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>7</td>\n",
       "      <td>79.162946</td>\n",
       "      <td>1.840006</td>\n",
       "      <td>23.967950</td>\n",
       "      <td>0.154458</td>\n",
       "      <td>10</td>\n",
       "      <td>scale</td>\n",
       "      <td>rbf</td>\n",
       "      <td>{'C': 10, 'gamma': 'scale', 'kernel': 'rbf'}</td>\n",
       "      <td>0.825840</td>\n",
       "      <td>...</td>\n",
       "      <td>0.003745</td>\n",
       "      <td>5</td>\n",
       "      <td>0.782990</td>\n",
       "      <td>0.795335</td>\n",
       "      <td>0.785045</td>\n",
       "      <td>0.793050</td>\n",
       "      <td>0.772895</td>\n",
       "      <td>0.785864</td>\n",
       "      <td>0.007980</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 25 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    Unnamed: 0  mean_fit_time  std_fit_time  mean_score_time  std_score_time  \\\n",
       "23          23     136.975841      2.200803        22.996254        0.472610   \n",
       "19          19     120.700509      1.970945        22.701440        0.465616   \n",
       "15          15     104.998210      1.040820        23.312394        0.183782   \n",
       "11          11      88.963005      1.429340        23.226407        0.503023   \n",
       "7            7      79.162946      1.840006        23.967950        0.154458   \n",
       "\n",
       "    param_C param_gamma param_kernel  \\\n",
       "23      100       scale          rbf   \n",
       "19       75       scale          rbf   \n",
       "15       50       scale          rbf   \n",
       "11       25       scale          rbf   \n",
       "7        10       scale          rbf   \n",
       "\n",
       "                                           params  split0_test_accuracy  ...  \\\n",
       "23  {'C': 100, 'gamma': 'scale', 'kernel': 'rbf'}              0.830246  ...   \n",
       "19   {'C': 75, 'gamma': 'scale', 'kernel': 'rbf'}              0.829500  ...   \n",
       "15   {'C': 50, 'gamma': 'scale', 'kernel': 'rbf'}              0.828305  ...   \n",
       "11   {'C': 25, 'gamma': 'scale', 'kernel': 'rbf'}              0.827035  ...   \n",
       "7    {'C': 10, 'gamma': 'scale', 'kernel': 'rbf'}              0.825840  ...   \n",
       "\n",
       "    std_test_accuracy  rank_test_accuracy  split0_test_F1  split1_test_F1  \\\n",
       "23           0.003774                   1        0.788064        0.799798   \n",
       "19           0.003748                   2        0.786916        0.798719   \n",
       "15           0.003650                   3        0.785875        0.797594   \n",
       "11           0.003642                   4        0.785165        0.797199   \n",
       "7            0.003745                   5        0.782990        0.795335   \n",
       "\n",
       "    split2_test_F1  split3_test_F1  split4_test_F1  mean_test_F1  std_test_F1  \\\n",
       "23        0.791047        0.793089        0.775469      0.789494     0.008003   \n",
       "19        0.788953        0.792964        0.775946      0.788700     0.007543   \n",
       "15        0.788175        0.793475        0.775783      0.788181     0.007425   \n",
       "11        0.786614        0.794482        0.774684      0.787629     0.007915   \n",
       "7         0.785045        0.793050        0.772895      0.785864     0.007980   \n",
       "\n",
       "    rank_test_F1  \n",
       "23             1  \n",
       "19             2  \n",
       "15             3  \n",
       "11             4  \n",
       "7              5  \n",
       "\n",
       "[5 rows x 25 columns]"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dr.sort_values(by=[\"mean_test_F1\"], ascending=False).head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/Ghosty/.local/lib/python3.7/site-packages/ipykernel_launcher.py:1: UserWarning: Boolean Series key will be reindexed to match DataFrame index.\n",
      "  \"\"\"Entry point for launching an IPython kernel.\n"
     ]
    }
   ],
   "source": [
    "Y_1 = [grid(dr,n,'0.001') for n in np.unique(dd[\"param_C\"].values)]\n",
    "Y_4 = [grid(dr,n,\"scale\") for n in np.unique(dd[\"param_C\"].values)]\n",
    "Y_2 = [grid(dr,n,'0.01') for n in np.unique(dd[\"param_C\"].values)]\n",
    "Y_3 = [grid(dr,n,'0.1') for n in np.unique(dd[\"param_C\"].values)]\n",
    "Z2 = np.vstack((Y_1, Y_2, Y_3, Y_4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [],
   "source": [
    "pgamma = np.array([0.001, 0.01, 0.1, 0.27] )\n",
    "\n",
    "C, G = np.meshgrid(np.unique(dd[\"param_C\"].values), pgamma)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10,7))\n",
    "plt.contourf(C, G, Z2, 20, cmap='RdGy')\n",
    "plt.colorbar()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.07008767133336684"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "1/(np.var(X_train)*X_train.shape[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Best params \n",
    "## DeepDense VAE"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = np.load(\"data/encoded_data_train_DeepDenseVAE_mark_V_[160-256-128-15]_VDO_GNLL_400ep_lr-1e-4.npy\")\n",
    "y_train = np.load(\"data/labels_train.npy\")\n",
    "X_test = np.load(\"data/encoded_data_test_DeepDenseVAE_mark_V_[160-256-128-15]_VDO_GNLL_400ep_lr-1e-4.npy\")\n",
    "y_test = np.load(\"data/labels_test.npy\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "dd_clf = SVC(kernel=\"rbf\", C=10, gamma=\"scale\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SVC(C=10, cache_size=200, class_weight=None, coef0=0.0,\n",
       "    decision_function_shape='ovr', degree=3, gamma='scale', kernel='rbf',\n",
       "    max_iter=-1, probability=False, random_state=None, shrinking=True,\n",
       "    tol=0.001, verbose=False)"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dd_clf.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['models_and_losses/SVM_DeepDenseVAE{kerner=rbf,C=10,gamma=scale=0.07}.joblib']"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#joblib.dump(dd_clf, \"models_and_losses/SVM_DeepDenseVAE{kerner=rbf,C=10,gamma=scale=0.07}.joblib\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred_dd = dd_clf.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8029322574632699"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "f1_score(y_true=y_test, y_pred = y_pred_dd, average=\"macro\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8417184512428298"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(y_true=y_test, y_pred = y_pred_dd)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[2567, 1520,   27,  121],\n",
       "       [ 341, 9948,   33,  123],\n",
       "       [  27,   35,  397,   20],\n",
       "       [ 222,  169,   11, 1175]])"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_matrix(y_true=y_test, y_pred = y_pred_dd)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## CNN-VAE"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = np.load(\"data/encoded_data_train_DeepConvVAE_ELU_NLL_250ep_lr-{1e-3, 1e-4}_bs-300.npy\")\n",
    "X_test = np.load(\"data/encoded_data_test_DeepConvVAE_ELU_NLL_250ep_lr-{1e-3, 1e-4}_bs-300.npy\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.07350118567710469"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "1/(np.var(X_train)*X_train.shape[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "dc_clf = SVC(kernel=\"rbf\", C=1, gamma=0.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SVC(C=1, cache_size=200, class_weight=None, coef0=0.0,\n",
       "    decision_function_shape='ovr', degree=3, gamma=0.1, kernel='rbf',\n",
       "    max_iter=-1, probability=False, random_state=None, shrinking=True,\n",
       "    tol=0.001, verbose=False)"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dc_clf.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['models_and_losses/SVM_CNN-VAE{kerner=rbf,C=1,gamma=0.1}.joblib']"
      ]
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#joblib.dump(dc_clf, \"models_and_losses/SVM_CNN-VAE{kerner=rbf,C=1,gamma=0.1}.joblib\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred_dc = dc_clf.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8074439550968406"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "f1_score(y_true=y_test, y_pred = y_pred_dc, average=\"macro\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8454827915869981"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(y_true=y_test, y_pred = y_pred_dc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 2542,  1538,    30,   125],\n",
       "       [  300, 10011,    31,   103],\n",
       "       [   21,    41,   399,    18],\n",
       "       [  201,   165,    13,  1198]])"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_matrix(y_true=y_test, y_pred = y_pred_dc)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## LSTM-VAE"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train1 = np.load(\"data/mu_lstm1.npy\")\n",
    "X_train2 = np.load(\"data/mu_lstm2.npy\")\n",
    "X_train3 = np.load(\"data/mu_lstm3.npy\")\n",
    "\n",
    "X_train = np.vstack((X_train1, X_train2, X_train3))\n",
    "\n",
    "X_test = np.load(\"data/mu_lstm_test.npy\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.2772047392196986"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "1/(np.var(X_train)*X_train.shape[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "dr_clf = SVC(kernel=\"rbf\", C=100, gamma=\"scale\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SVC(C=100, cache_size=200, class_weight=None, coef0=0.0,\n",
       "    decision_function_shape='ovr', degree=3, gamma='scale', kernel='rbf',\n",
       "    max_iter=-1, probability=False, random_state=None, shrinking=True,\n",
       "    tol=0.001, verbose=False)"
      ]
     },
     "execution_count": 78,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dr_clf.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['models_and_losses/SVM_LSTM-VAE{kerner=rbf,C=100,gamma=scale=0.27}.joblib']"
      ]
     },
     "execution_count": 83,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#joblib.dump(dc_clf, \"models_and_losses/SVM_LSTM-VAE{kerner=rbf,C=100,gamma=scale=0.27}.joblib\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred_dr = dr_clf.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.803976007985282"
      ]
     },
     "execution_count": 80,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "f1_score(y_true=y_test, y_pred = y_pred_dr, average=\"macro\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8381931166347992"
      ]
     },
     "execution_count": 81,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(y_true=y_test, y_pred = y_pred_dr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 2367,  1722,    26,   120],\n",
       "       [  296, 10013,    41,    95],\n",
       "       [   22,    34,   404,    19],\n",
       "       [  183,   137,    13,  1244]])"
      ]
     },
     "execution_count": 82,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_matrix(y_true=y_test, y_pred = y_pred_dr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Pytorch",
   "language": "python",
   "name": "pytorch"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}