Stable Compounds, Random Forest GridSearch.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#  Stable Compounds, Proof of Concept"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Imports and Data Loading"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import OneHotEncoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.metrics import classification_report\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.multiclass import OneVsRestClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.svm import SVC\n",
    "from sklearn.linear_model import LinearRegression\n",
    "from sklearn.model_selection import GridSearchCV, cross_validate\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn.metrics import recall_score, precision_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.decomposition import PCA"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Steps\n",
    "- Data Ingestion and EDA\n",
    "- Initial Model Construction\n",
    "- Model Refinement\n",
    "- Model Evaluation (Model Choice)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "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>formulaA</th>\n",
       "      <th>formulaB</th>\n",
       "      <th>formulaA_elements_AtomicVolume</th>\n",
       "      <th>formulaB_elements_AtomicVolume</th>\n",
       "      <th>formulaA_elements_AtomicWeight</th>\n",
       "      <th>formulaB_elements_AtomicWeight</th>\n",
       "      <th>formulaA_elements_BoilingT</th>\n",
       "      <th>formulaB_elements_BoilingT</th>\n",
       "      <th>formulaA_elements_BulkModulus</th>\n",
       "      <th>formulaB_elements_BulkModulus</th>\n",
       "      <th>...</th>\n",
       "      <th>formulaB_elements_Row</th>\n",
       "      <th>formulaA_elements_ShearModulus</th>\n",
       "      <th>formulaB_elements_ShearModulus</th>\n",
       "      <th>formulaA_elements_SpaceGroupNumber</th>\n",
       "      <th>formulaB_elements_SpaceGroupNumber</th>\n",
       "      <th>avg_coordination_A</th>\n",
       "      <th>avg_coordination_B</th>\n",
       "      <th>avg_nearest_neighbor_distance_A</th>\n",
       "      <th>avg_nearest_neighbor_distance_B</th>\n",
       "      <th>stabilityVec</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Ac</td>\n",
       "      <td>Ag</td>\n",
       "      <td>37.433086</td>\n",
       "      <td>17.075648</td>\n",
       "      <td>227.000</td>\n",
       "      <td>107.868200</td>\n",
       "      <td>3473.0</td>\n",
       "      <td>2435.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>100.0</td>\n",
       "      <td>...</td>\n",
       "      <td>5</td>\n",
       "      <td>0.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>225</td>\n",
       "      <td>225</td>\n",
       "      <td>12.0</td>\n",
       "      <td>12.0</td>\n",
       "      <td>3.99462</td>\n",
       "      <td>2.94195</td>\n",
       "      <td>[1.0,0.0,0.0,1.0,0.0,1.0,0.0,0.0,0.0,0.0,1.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Ac</td>\n",
       "      <td>Al</td>\n",
       "      <td>37.433086</td>\n",
       "      <td>16.594425</td>\n",
       "      <td>227.000</td>\n",
       "      <td>26.981539</td>\n",
       "      <td>3473.0</td>\n",
       "      <td>2792.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>76.0</td>\n",
       "      <td>...</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0</td>\n",
       "      <td>26.0</td>\n",
       "      <td>225</td>\n",
       "      <td>225</td>\n",
       "      <td>12.0</td>\n",
       "      <td>12.0</td>\n",
       "      <td>3.99462</td>\n",
       "      <td>2.85595</td>\n",
       "      <td>[1.0,0.0,0.0,1.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Ac</td>\n",
       "      <td>As</td>\n",
       "      <td>37.433086</td>\n",
       "      <td>21.723966</td>\n",
       "      <td>227.000</td>\n",
       "      <td>74.921600</td>\n",
       "      <td>3473.0</td>\n",
       "      <td>887.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>22.0</td>\n",
       "      <td>...</td>\n",
       "      <td>4</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>225</td>\n",
       "      <td>166</td>\n",
       "      <td>12.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>3.99462</td>\n",
       "      <td>2.55790</td>\n",
       "      <td>[1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Ac</td>\n",
       "      <td>Ba</td>\n",
       "      <td>37.433086</td>\n",
       "      <td>64.969282</td>\n",
       "      <td>227.000</td>\n",
       "      <td>137.327000</td>\n",
       "      <td>3473.0</td>\n",
       "      <td>2143.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>9.6</td>\n",
       "      <td>...</td>\n",
       "      <td>6</td>\n",
       "      <td>0.0</td>\n",
       "      <td>4.9</td>\n",
       "      <td>225</td>\n",
       "      <td>229</td>\n",
       "      <td>12.0</td>\n",
       "      <td>8.0</td>\n",
       "      <td>3.99462</td>\n",
       "      <td>4.35637</td>\n",
       "      <td>[1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Ac</td>\n",
       "      <td>Bi</td>\n",
       "      <td>37.433086</td>\n",
       "      <td>35.483459</td>\n",
       "      <td>227.000</td>\n",
       "      <td>208.980400</td>\n",
       "      <td>3473.0</td>\n",
       "      <td>1837.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>31.0</td>\n",
       "      <td>...</td>\n",
       "      <td>6</td>\n",
       "      <td>0.0</td>\n",
       "      <td>12.0</td>\n",
       "      <td>225</td>\n",
       "      <td>12</td>\n",
       "      <td>12.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>3.99462</td>\n",
       "      <td>3.11221</td>\n",
       "      <td>[1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]</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>2567</th>\n",
       "      <td>Zr</td>\n",
       "      <td>Th</td>\n",
       "      <td>23.265943</td>\n",
       "      <td>32.865683</td>\n",
       "      <td>91.224</td>\n",
       "      <td>232.038060</td>\n",
       "      <td>4682.0</td>\n",
       "      <td>5093.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>54.0</td>\n",
       "      <td>...</td>\n",
       "      <td>7</td>\n",
       "      <td>33.0</td>\n",
       "      <td>31.0</td>\n",
       "      <td>194</td>\n",
       "      <td>225</td>\n",
       "      <td>12.0</td>\n",
       "      <td>12.0</td>\n",
       "      <td>3.19147</td>\n",
       "      <td>3.56059</td>\n",
       "      <td>[1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2568</th>\n",
       "      <td>Zr</td>\n",
       "      <td>Tl</td>\n",
       "      <td>23.265943</td>\n",
       "      <td>28.640877</td>\n",
       "      <td>91.224</td>\n",
       "      <td>204.383300</td>\n",
       "      <td>4682.0</td>\n",
       "      <td>1746.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>43.0</td>\n",
       "      <td>...</td>\n",
       "      <td>6</td>\n",
       "      <td>33.0</td>\n",
       "      <td>2.8</td>\n",
       "      <td>194</td>\n",
       "      <td>194</td>\n",
       "      <td>12.0</td>\n",
       "      <td>8.0</td>\n",
       "      <td>3.19147</td>\n",
       "      <td>3.43253</td>\n",
       "      <td>[1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0,0.0,1.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2569</th>\n",
       "      <td>Zr</td>\n",
       "      <td>V</td>\n",
       "      <td>23.265943</td>\n",
       "      <td>13.844898</td>\n",
       "      <td>91.224</td>\n",
       "      <td>50.941500</td>\n",
       "      <td>4682.0</td>\n",
       "      <td>3680.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>160.0</td>\n",
       "      <td>...</td>\n",
       "      <td>4</td>\n",
       "      <td>33.0</td>\n",
       "      <td>47.0</td>\n",
       "      <td>194</td>\n",
       "      <td>229</td>\n",
       "      <td>12.0</td>\n",
       "      <td>8.0</td>\n",
       "      <td>3.19147</td>\n",
       "      <td>2.59229</td>\n",
       "      <td>[1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2570</th>\n",
       "      <td>Zr</td>\n",
       "      <td>Xe</td>\n",
       "      <td>23.265943</td>\n",
       "      <td>36952.924020</td>\n",
       "      <td>91.224</td>\n",
       "      <td>131.293000</td>\n",
       "      <td>4682.0</td>\n",
       "      <td>165.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>5</td>\n",
       "      <td>33.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>194</td>\n",
       "      <td>225</td>\n",
       "      <td>12.0</td>\n",
       "      <td>12.0</td>\n",
       "      <td>3.19147</td>\n",
       "      <td>4.85032</td>\n",
       "      <td>[1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2571</th>\n",
       "      <td>Zr</td>\n",
       "      <td>Zn</td>\n",
       "      <td>23.265943</td>\n",
       "      <td>15.205684</td>\n",
       "      <td>91.224</td>\n",
       "      <td>65.380000</td>\n",
       "      <td>4682.0</td>\n",
       "      <td>1180.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>70.0</td>\n",
       "      <td>...</td>\n",
       "      <td>4</td>\n",
       "      <td>33.0</td>\n",
       "      <td>43.0</td>\n",
       "      <td>194</td>\n",
       "      <td>194</td>\n",
       "      <td>12.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>3.19147</td>\n",
       "      <td>2.62673</td>\n",
       "      <td>[1.0,0.0,0.0,1.0,0.0,1.0,0.0,0.0,0.0,0.0,1.0]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2572 rows × 99 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     formulaA formulaB  formulaA_elements_AtomicVolume  \\\n",
       "0          Ac       Ag                       37.433086   \n",
       "1          Ac       Al                       37.433086   \n",
       "2          Ac       As                       37.433086   \n",
       "3          Ac       Ba                       37.433086   \n",
       "4          Ac       Bi                       37.433086   \n",
       "...       ...      ...                             ...   \n",
       "2567       Zr       Th                       23.265943   \n",
       "2568       Zr       Tl                       23.265943   \n",
       "2569       Zr        V                       23.265943   \n",
       "2570       Zr       Xe                       23.265943   \n",
       "2571       Zr       Zn                       23.265943   \n",
       "\n",
       "      formulaB_elements_AtomicVolume  formulaA_elements_AtomicWeight  \\\n",
       "0                          17.075648                         227.000   \n",
       "1                          16.594425                         227.000   \n",
       "2                          21.723966                         227.000   \n",
       "3                          64.969282                         227.000   \n",
       "4                          35.483459                         227.000   \n",
       "...                              ...                             ...   \n",
       "2567                       32.865683                          91.224   \n",
       "2568                       28.640877                          91.224   \n",
       "2569                       13.844898                          91.224   \n",
       "2570                    36952.924020                          91.224   \n",
       "2571                       15.205684                          91.224   \n",
       "\n",
       "      formulaB_elements_AtomicWeight  formulaA_elements_BoilingT  \\\n",
       "0                         107.868200                      3473.0   \n",
       "1                          26.981539                      3473.0   \n",
       "2                          74.921600                      3473.0   \n",
       "3                         137.327000                      3473.0   \n",
       "4                         208.980400                      3473.0   \n",
       "...                              ...                         ...   \n",
       "2567                      232.038060                      4682.0   \n",
       "2568                      204.383300                      4682.0   \n",
       "2569                       50.941500                      4682.0   \n",
       "2570                      131.293000                      4682.0   \n",
       "2571                       65.380000                      4682.0   \n",
       "\n",
       "      formulaB_elements_BoilingT  formulaA_elements_BulkModulus  \\\n",
       "0                         2435.0                            0.0   \n",
       "1                         2792.0                            0.0   \n",
       "2                          887.0                            0.0   \n",
       "3                         2143.0                            0.0   \n",
       "4                         1837.0                            0.0   \n",
       "...                          ...                            ...   \n",
       "2567                      5093.0                            0.0   \n",
       "2568                      1746.0                            0.0   \n",
       "2569                      3680.0                            0.0   \n",
       "2570                       165.0                            0.0   \n",
       "2571                      1180.0                            0.0   \n",
       "\n",
       "      formulaB_elements_BulkModulus  ...  formulaB_elements_Row  \\\n",
       "0                             100.0  ...                      5   \n",
       "1                              76.0  ...                      3   \n",
       "2                              22.0  ...                      4   \n",
       "3                               9.6  ...                      6   \n",
       "4                              31.0  ...                      6   \n",
       "...                             ...  ...                    ...   \n",
       "2567                           54.0  ...                      7   \n",
       "2568                           43.0  ...                      6   \n",
       "2569                          160.0  ...                      4   \n",
       "2570                            0.0  ...                      5   \n",
       "2571                           70.0  ...                      4   \n",
       "\n",
       "      formulaA_elements_ShearModulus  formulaB_elements_ShearModulus  \\\n",
       "0                                0.0                            30.0   \n",
       "1                                0.0                            26.0   \n",
       "2                                0.0                             0.0   \n",
       "3                                0.0                             4.9   \n",
       "4                                0.0                            12.0   \n",
       "...                              ...                             ...   \n",
       "2567                            33.0                            31.0   \n",
       "2568                            33.0                             2.8   \n",
       "2569                            33.0                            47.0   \n",
       "2570                            33.0                             0.0   \n",
       "2571                            33.0                            43.0   \n",
       "\n",
       "      formulaA_elements_SpaceGroupNumber  formulaB_elements_SpaceGroupNumber  \\\n",
       "0                                    225                                 225   \n",
       "1                                    225                                 225   \n",
       "2                                    225                                 166   \n",
       "3                                    225                                 229   \n",
       "4                                    225                                  12   \n",
       "...                                  ...                                 ...   \n",
       "2567                                 194                                 225   \n",
       "2568                                 194                                 194   \n",
       "2569                                 194                                 229   \n",
       "2570                                 194                                 225   \n",
       "2571                                 194                                 194   \n",
       "\n",
       "      avg_coordination_A  avg_coordination_B  avg_nearest_neighbor_distance_A  \\\n",
       "0                   12.0                12.0                          3.99462   \n",
       "1                   12.0                12.0                          3.99462   \n",
       "2                   12.0                 3.0                          3.99462   \n",
       "3                   12.0                 8.0                          3.99462   \n",
       "4                   12.0                 3.0                          3.99462   \n",
       "...                  ...                 ...                              ...   \n",
       "2567                12.0                12.0                          3.19147   \n",
       "2568                12.0                 8.0                          3.19147   \n",
       "2569                12.0                 8.0                          3.19147   \n",
       "2570                12.0                12.0                          3.19147   \n",
       "2571                12.0                 6.0                          3.19147   \n",
       "\n",
       "      avg_nearest_neighbor_distance_B  \\\n",
       "0                             2.94195   \n",
       "1                             2.85595   \n",
       "2                             2.55790   \n",
       "3                             4.35637   \n",
       "4                             3.11221   \n",
       "...                               ...   \n",
       "2567                          3.56059   \n",
       "2568                          3.43253   \n",
       "2569                          2.59229   \n",
       "2570                          4.85032   \n",
       "2571                          2.62673   \n",
       "\n",
       "                                       stabilityVec  \n",
       "0     [1.0,0.0,0.0,1.0,0.0,1.0,0.0,0.0,0.0,0.0,1.0]  \n",
       "1     [1.0,0.0,0.0,1.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]  \n",
       "2     [1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]  \n",
       "3     [1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]  \n",
       "4     [1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]  \n",
       "...                                             ...  \n",
       "2567  [1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]  \n",
       "2568  [1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0,0.0,1.0]  \n",
       "2569  [1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]  \n",
       "2570  [1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]  \n",
       "2571  [1.0,0.0,0.0,1.0,0.0,1.0,0.0,0.0,0.0,0.0,1.0]  \n",
       "\n",
       "[2572 rows x 99 columns]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "my_path = './challenge_data/training_data.csv'\n",
    "df = pd.read_csv(my_path)\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "my_path2 = './challenge_data/test_data.csv'\n",
    "test = pd.read_csv(my_path2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Ingestion and EDA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# df.iloc[:,:10].hist(figsize=(12,10),bins=50);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# len(pd.unique(df['formulaB']))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Removal of Stability Vector and Features relating to positioning in element table as many other properties are defining "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature=df.drop(['stabilityVec'], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature2=feature.drop(['formulaA', 'formulaB','formulaA_elements_Column','formulaB_elements_Column',\n",
    "                      'formulaA_elements_MendeleevNumber','formulaB_elements_MendeleevNumber',\n",
    "                      'formulaA_elements_Number','formulaB_elements_Number',\n",
    "                      'formulaA_elements_Row','formulaB_elements_Row'], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "test2=test.drop(['formulaA', 'formulaB','formulaA_elements_Column','formulaB_elements_Column',\n",
    "                      'formulaA_elements_MendeleevNumber','formulaB_elements_MendeleevNumber',\n",
    "                      'formulaA_elements_Number','formulaB_elements_Number',\n",
    "                      'formulaA_elements_Row','formulaB_elements_Row'], axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Creation of a heatmap of features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 792x648 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#cell taken from class notebook\n",
    "# extract the first 20 features\n",
    "first_twenty = feature2.iloc[:, :20] \n",
    "# compute correlation matrix\n",
    "corr = first_twenty.corr()  \n",
    "# make mask\n",
    "mask = np.zeros_like(corr, dtype=np.bool)  \n",
    "# mask the upper triangle\n",
    "mask[np.triu_indices_from(mask)] = True  \n",
    "\n",
    "# create a figure and a subplot\n",
    "fig, ax = plt.subplots(figsize=(11, 9))  \n",
    "# costumizing the color map\n",
    "cmap = sns.diverging_palette(220, 10, as_cmap=True) \n",
    "sns.heatmap(\n",
    "    corr,\n",
    "    mask=mask,\n",
    "    cmap=cmap,\n",
    "    center=0,\n",
    "    linewidth=0.5,\n",
    "    cbar_kws={'shrink': 0.5}\n",
    ");"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Extraction of the target , conversion to float and into a numpy array"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "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>0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[1.0,0.0,0.0,1.0,0.0,1.0,0.0,0.0,0.0,0.0,1.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>[1.0,0.0,0.0,1.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>[1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                               0\n",
       "0  [1.0,0.0,0.0,1.0,0.0,1.0,0.0,0.0,0.0,0.0,1.0]\n",
       "1  [1.0,0.0,0.0,1.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]\n",
       "2  [1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "target=df[['stabilityVec']]\n",
    "target.columns = range(target.shape[1])\n",
    "target.iloc[:3,:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "b=np.zeros((1,11))\n",
    "for idx, row in target.iterrows():\n",
    "    ser=target.iat[idx,0]\n",
    "    ser=pd.DataFrame(list(ser.split(\",\")) )\n",
    "    ser=ser.T\n",
    "    a=np.zeros((1,11))\n",
    "    for i,col in enumerate(ser):\n",
    "        elem=ser.iat[0,i]\n",
    "        if i==0:\n",
    "            new=float(elem[1:4])\n",
    "        if i==10:\n",
    "            new=float(elem[0:3])\n",
    "        if i>0 and i<10:\n",
    "            new=float(elem)\n",
    "        a[0,i]=new\n",
    "    b=np.vstack((b, a))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[1., 0., 0., 1., 0., 1., 0., 0., 0., 0., 1.],\n",
       "       [1., 0., 0., 1., 0., 0., 0., 0., 0., 0., 1.],\n",
       "       [1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.]])"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "target_np=b[1:]\n",
    "target_np[:3]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Distribution of class labels training data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_pos = [0,1,2,3,4,5,6,7,8,9,10]#np.arange(11)\n",
    "squares = [np.count_nonzero(x) for x in target_np.T]\n",
    "\n",
    "plt.bar(y_pos, squares, align='center', alpha=0.5)\n",
    "plt.xticks(y_pos)\n",
    "plt.ylabel('# Compund Mixtures')\n",
    "plt.title('Frequency of Stable Compound Mixtures Ratios')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Functions for counting nonzero samples, and total stable mixed compounds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "def nonzero_calc(target):  \n",
    "    target_min=target[:,1:10]\n",
    "\n",
    "    nonzs_rowsT=[]\n",
    "    for rowT in target_min:\n",
    "        nonz_rowT=(rowT != 0).sum()\n",
    "        nonzs_rowsT.append(nonz_rowT)\n",
    "    # print(nonzs_rowsT[:1000])\n",
    "#     print(np.count_nonzero(nonzs_rowsT))\n",
    "    return np.count_nonzero(nonzs_rowsT),sum(nonzs_rowsT)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1228, 2493)"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nonzero_calc(target_np)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "One-Hot Encoding of Geometry features on complete Dataset (train_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "comb1=pd.concat((feature2,test2), axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "comb_enc1=comb1[['formulaA_elements_SpaceGroupNumber','formulaB_elements_SpaceGroupNumber',\n",
    "                 'avg_coordination_A','avg_coordination_B']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "ohe = OneHotEncoder(sparse = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\ProgramData\\Miniconda3\\lib\\site-packages\\sklearn\\preprocessing\\_encoders.py:415: FutureWarning: The handling of integer data will change in version 0.22. Currently, the categories are determined based on the range [0, max(values)], while in the future they will be determined based on the unique values.\n",
      "If you want the future behaviour and silence this warning, you can specify \"categories='auto'\".\n",
      "In case you used a LabelEncoder before this OneHotEncoder to convert the categories to integers, then you can now use the OneHotEncoder directly.\n",
      "  warnings.warn(msg, FutureWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[0., 0., 0., ..., 0., 0., 1.],\n",
       "       [0., 0., 0., ..., 0., 0., 1.],\n",
       "       [0., 0., 0., ..., 0., 0., 0.],\n",
       "       ...,\n",
       "       [0., 0., 0., ..., 0., 0., 1.],\n",
       "       [0., 0., 0., ..., 0., 0., 1.],\n",
       "       [0., 0., 0., ..., 0., 1., 0.]])"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ohe.fit(comb_enc1)\n",
    "\n",
    "# Transforming the \"training\" data\n",
    "comb_encdd1 = ohe.transform(comb_enc1)\n",
    "comb_encdd1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[array([  2.,  12.,  14.,  62.,  63.,  64.,  70., 139., 141., 152., 166.,\n",
       "        194., 217., 225., 227., 229.]),\n",
       " array([  2.,  12.,  14.,  62.,  63.,  64.,  70., 139., 141., 152., 166.,\n",
       "        194., 217., 225., 227., 229.]),\n",
       " array([ 1.,  2.,  3.,  4.,  5.,  6.,  8., 12.]),\n",
       " array([ 1.,  2.,  3.,  4.,  5.,  6.,  8., 12.])]"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ohe.categories_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Removal and subsequent concatenation of binary features, standardization of remaining features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "# cat={}\n",
    "# for i,col in enumerate(comb1):\n",
    "#     num=len(comb1[col].unique())\n",
    "#     cat[i]=num\n",
    "# cat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "binaries=feature2.iloc[:,[42,43,44,45,46,47,48,49,50,51,52,53,74,75]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "bins=binaries.to_numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "bin_test=test2.iloc[:,[42,43,44,45,46,47,48,49,50,51,52,53,74,75]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "bin_test2=bin_test.to_numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature3=feature2.drop(feature2.columns[[42,43,44,45,46,47,48,49,50,51,52,53,74,75,82,83,84,85]],axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "test3=test2.drop(test2.columns[[42,43,44,45,46,47,48,49,50,51,52,53,74,75,82,83,84,85]],axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "scalar=StandardScaler()\n",
    "scalar.fit(feature3)\n",
    "X_tr=scalar.transform(feature3)\n",
    "X_ts=scalar.transform(test3)\n",
    "# print(X_tr)\n",
    "# print(X_ts)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_tr2=np.concatenate((X_tr,bins), axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_ts2=np.concatenate((X_ts,bin_test2), axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2572, 84)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_tr2.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_tr3=np.concatenate((X_tr2,comb_encdd1[:2572]), axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_ts3=np.concatenate((X_ts2,comb_encdd1[2572:3321]), axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Initial Model Construction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_val, y_train, y_val = train_test_split(X_tr3, target_np)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "!Going forward, use cross-validation instead of these functions## Function for calculating the probability difference of classes (certainty) on test set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "def prob_diff(pred_prob):   \n",
    "    differences=[]\n",
    "    for index,arr in enumerate(pred_prob):\n",
    "        if index==0: \n",
    "            continue\n",
    "        if index>9:\n",
    "            break  \n",
    "        if index>0 or index<10:\n",
    "            means=np.mean(arr[:,0]-arr[:,1])\n",
    "            differences.append(means)\n",
    "    return np.min(differences)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "def prob_diff2(pred_prob):\n",
    "    mins=[]\n",
    "    for index,col in enumerate(pred_prob.T):\n",
    "        if index==0: \n",
    "            continue\n",
    "        if index>9:\n",
    "            break  \n",
    "        if index>0 or index<10:\n",
    "            min_avg=np.mean(col)\n",
    "            mins.append(min_avg)\n",
    "    return 1-2*np.min(mins)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Random Forest Base Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Random Forest base model: \n",
      "               precision    recall  f1-score   support\n",
      "\n",
      "           0       1.00      1.00      1.00       643\n",
      "           1       1.00      0.06      0.12        16\n",
      "           2       0.67      0.08      0.14        25\n",
      "           3       0.76      0.65      0.70       150\n",
      "           4       0.62      0.23      0.33        35\n",
      "           5       0.77      0.60      0.68       141\n",
      "           6       0.73      0.20      0.31        40\n",
      "           7       0.80      0.41      0.55        80\n",
      "           8       0.78      0.47      0.59        92\n",
      "           9       0.83      0.38      0.53        13\n",
      "          10       1.00      1.00      1.00       643\n",
      "\n",
      "   micro avg       0.95      0.83      0.89      1878\n",
      "   macro avg       0.81      0.46      0.54      1878\n",
      "weighted avg       0.93      0.83      0.86      1878\n",
      " samples avg       0.96      0.89      0.91      1878\n",
      " \n",
      "\n",
      "Validation score: \n",
      " 0.6267496111975117\n",
      "Minimal Probability difference: 0.5148464619492656\n"
     ]
    }
   ],
   "source": [
    "rf_base = RandomForestClassifier(n_estimators=100)\n",
    "rf_base.fit(X_train, y_train)\n",
    "print(\"Random Forest base model: \\n\", classification_report(y_val, rf_base.predict(X_val)), \"\\n\")\n",
    "print(\"Validation score: \\n\",rf_base.score(X_val, y_val))\n",
    "print('Minimal Probability difference:', prob_diff(rf_base.predict_proba(X_ts3)))  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Multivariate Linear Regression Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Linear Regression: \n",
      "               precision    recall  f1-score   support\n",
      "\n",
      "           0       1.00      1.00      1.00       643\n",
      "           1       0.00      0.00      0.00        16\n",
      "           2       0.00      0.00      0.00        25\n",
      "           3       0.63      0.40      0.49       150\n",
      "           4       0.33      0.06      0.10        35\n",
      "           5       0.56      0.29      0.38       141\n",
      "           6       0.29      0.05      0.09        40\n",
      "           7       0.40      0.07      0.13        80\n",
      "           8       0.66      0.27      0.38        92\n",
      "           9       0.00      0.00      0.00        13\n",
      "          10       1.00      1.00      1.00       643\n",
      "\n",
      "   micro avg       0.94      0.76      0.84      1878\n",
      "   macro avg       0.44      0.29      0.32      1878\n",
      "weighted avg       0.84      0.76      0.78      1878\n",
      " samples avg       0.96      0.83      0.87      1878\n",
      " \n",
      "\n",
      "Validation score: \n",
      " 0.1816389102497435\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\ProgramData\\Miniconda3\\lib\\site-packages\\sklearn\\metrics\\classification.py:1437: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples.\n",
      "  'precision', 'predicted', average, warn_for)\n",
      "C:\\ProgramData\\Miniconda3\\lib\\site-packages\\sklearn\\base.py:420: FutureWarning: The default value of multioutput (not exposed in score method) will change from 'variance_weighted' to 'uniform_average' in 0.23 to keep consistent with 'metrics.r2_score'. To specify the default value manually and avoid the warning, please either call 'metrics.r2_score' directly or make a custom scorer with 'metrics.make_scorer' (the built-in scorer 'r2' uses multioutput='uniform_average').\n",
      "  \"multioutput='uniform_average').\", FutureWarning)\n"
     ]
    }
   ],
   "source": [
    "lr = LinearRegression()\n",
    "lr.fit(X_train,y_train)\n",
    "y_val_pred_cont=lr.predict(X_val)\n",
    "y_val_pred_cont\n",
    "y_val_pred=np.where(y_val_pred_cont < 0.5, 0.0, 1.0)\n",
    "y_val_pred\n",
    "print(\"Linear Regression: \\n\", classification_report(y_val, y_val_pred), \"\\n\")\n",
    "print(\"Validation score: \\n\",lr.score(X_val, y_val_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Support Vector Machine Model (OneVsRest Wrapper Used to enable multi-class target)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\ProgramData\\Miniconda3\\lib\\site-packages\\sklearn\\multiclass.py:76: UserWarning: Label 0 is present in all training examples.\n",
      "  str(classes[c]))\n",
      "C:\\ProgramData\\Miniconda3\\lib\\site-packages\\sklearn\\multiclass.py:76: UserWarning: Label 10 is present in all training examples.\n",
      "  str(classes[c]))\n",
      "C:\\ProgramData\\Miniconda3\\lib\\site-packages\\sklearn\\metrics\\classification.py:1437: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples.\n",
      "  'precision', 'predicted', average, warn_for)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Support Vector Machine: \n",
      "               precision    recall  f1-score   support\n",
      "\n",
      "           0       1.00      1.00      1.00       643\n",
      "           1       0.00      0.00      0.00        16\n",
      "           2       0.00      0.00      0.00        25\n",
      "           3       0.80      0.37      0.51       150\n",
      "           4       0.00      0.00      0.00        35\n",
      "           5       0.82      0.22      0.35       141\n",
      "           6       0.00      0.00      0.00        40\n",
      "           7       0.00      0.00      0.00        80\n",
      "           8       0.50      0.03      0.06        92\n",
      "           9       0.00      0.00      0.00        13\n",
      "          10       1.00      1.00      1.00       643\n",
      "\n",
      "   micro avg       0.98      0.73      0.84      1878\n",
      "   macro avg       0.37      0.24      0.27      1878\n",
      "weighted avg       0.83      0.73      0.75      1878\n",
      " samples avg       0.99      0.82      0.88      1878\n",
      " \n",
      "\n",
      "Validation score: \n",
      " 0.5629860031104199\n",
      "Minimal Probability difference: 0.9671174181198008\n"
     ]
    }
   ],
   "source": [
    "clf = OneVsRestClassifier(SVC(gamma='auto',probability=True)).fit(X_train, y_train)\n",
    "print(\"Support Vector Machine: \\n\", classification_report(y_val, clf.predict(X_val)), \"\\n\")\n",
    "print(\"Validation score: \\n\",clf.score(X_val, y_val))\n",
    "print('Minimal Probability difference:',prob_diff2(clf.predict_proba(X_ts3)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "k-Nearest Neighbor Model (Algorithm Choice)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "KNeighbors: \n",
      "               precision    recall  f1-score   support\n",
      "\n",
      "           0       1.00      1.00      1.00       643\n",
      "           1       0.50      0.06      0.11        16\n",
      "           2       0.67      0.16      0.26        25\n",
      "           3       0.73      0.69      0.71       150\n",
      "           4       0.39      0.26      0.31        35\n",
      "           5       0.69      0.61      0.65       141\n",
      "           6       0.45      0.23      0.30        40\n",
      "           7       0.58      0.42      0.49        80\n",
      "           8       0.61      0.45      0.52        92\n",
      "           9       0.40      0.15      0.22        13\n",
      "          10       1.00      1.00      1.00       643\n",
      "\n",
      "   micro avg       0.91      0.84      0.87      1878\n",
      "   macro avg       0.64      0.46      0.51      1878\n",
      "weighted avg       0.88      0.84      0.85      1878\n",
      " samples avg       0.94      0.89      0.90      1878\n",
      " \n",
      "\n",
      "Validation score: \n",
      " 0.5800933125972006\n",
      "Minimal Probability difference: 0.5359145527369826\n"
     ]
    }
   ],
   "source": [
    "knn = KNeighborsClassifier()\n",
    "knn.fit(X_train, y_train)\n",
    "print(\"KNeighbors: \\n\", classification_report(y_val, knn.predict(X_val)), \"\\n\")\n",
    "print(\"Validation score: \\n\",knn.score(X_val, y_val))\n",
    "print('Minimal Probability difference:', prob_diff(knn.predict_proba(X_ts3)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Alternative: fitting all methods in a series:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "# estimators = [\n",
    "#     ('Random Forest - base model',\n",
    "#      RandomForestClassifier(n_estimators=100),False),\n",
    "\n",
    "#     ('Linear regression',\n",
    "#      LinearRegression(),True),\n",
    "\n",
    "#     ('Support Vector Machine',\n",
    "#      OneVsRestClassifier(SVC(gamma='auto',probability=True)),False),\n",
    "\n",
    "#     ('k-Nearest Neighbor',\n",
    "#      KNeighborsClassifier(),False)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "# for name, est, y_process in estimators:\n",
    "#     est.fit(X_train,y_train)\n",
    "#     if y_process==True:\n",
    "#         y_pred=est.predict(X_val)\n",
    "#         y_pred=np.where(y_pred < 0.5, 0.0, 1.0)\n",
    "#     if y_process==False:\n",
    "#         y_pred=est.predict(X_val)\n",
    "#     print(name,\"\\n\", classification_report(y_val,y_pred), \"\\n\")\n",
    "#     print(\"Validation score: \\n\",est.score(X_val, y_val))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Principal Component Analysis (PCA)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x2ded5f88688>]"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "n_samples,n_features= X_tr3.shape[0], X_tr3.shape[1]\n",
    "fig = plt.figure()\n",
    "ax = plt.axes()\n",
    "vs=[]\n",
    "expls=[]\n",
    "for v in range(0,min(n_samples, n_features)):\n",
    "    pca = PCA(n_components=v)\n",
    "    pca.fit(X_tr3)\n",
    "    expl=pca.explained_variance_ratio_  \n",
    "    vs.append(v)\n",
    "    expls.append(expl.sum())\n",
    "ax.plot(vs, expls)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model Refinement"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Random Forest Refinements"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Optimizing features and trees"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 1min 14s\n"
     ]
    },
    {
     "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>recall</th>\n",
       "      <th>trees</th>\n",
       "      <th>feat</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.439584</td>\n",
       "      <td>50</td>\n",
       "      <td>auto</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.464511</td>\n",
       "      <td>75</td>\n",
       "      <td>auto</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.451491</td>\n",
       "      <td>100</td>\n",
       "      <td>auto</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.452861</td>\n",
       "      <td>125</td>\n",
       "      <td>auto</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.454538</td>\n",
       "      <td>150</td>\n",
       "      <td>auto</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.445047</td>\n",
       "      <td>50</td>\n",
       "      <td>132</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.450508</td>\n",
       "      <td>75</td>\n",
       "      <td>132</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.446691</td>\n",
       "      <td>100</td>\n",
       "      <td>132</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.44923</td>\n",
       "      <td>125</td>\n",
       "      <td>132</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.453469</td>\n",
       "      <td>150</td>\n",
       "      <td>132</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     recall trees  feat\n",
       "0  0.439584    50  auto\n",
       "1  0.464511    75  auto\n",
       "2  0.451491   100  auto\n",
       "3  0.452861   125  auto\n",
       "4  0.454538   150  auto\n",
       "5  0.445047    50   132\n",
       "6  0.450508    75   132\n",
       "7  0.446691   100   132\n",
       "8   0.44923   125   132\n",
       "9  0.453469   150   132"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "num_features = ['auto', 132]\n",
    "n_estimators = [50,75,100,125,150]\n",
    "scores = dict()\n",
    "i = 0\n",
    "for f in num_features:\n",
    "    for e in n_estimators:\n",
    "        rf_est = RandomForestClassifier(n_estimators = e, max_features = f)#, random_state = 1738)\n",
    "        rf_est.fit(X_train, y_train)\n",
    "        scores[i] = {'recall':recall_score(y_val, rf_est.predict(X_val),average='macro'), 'trees' :e, \"feat\":f}\n",
    "        i+=1\n",
    "pd.DataFrame(scores).T"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "GridSearch for optimal weights"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Credit: https://www.kaggle.com/eikedehling/exploring-class-imbalance-resampling-and-weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\ProgramData\\Miniconda3\\lib\\site-packages\\sklearn\\model_selection\\_split.py:1978: FutureWarning: The default value of cv will change from 3 to 5 in version 0.22. Specify it explicitly to silence this warning.\n",
      "  warnings.warn(CV_WARNING, FutureWarning)\n",
      "C:\\ProgramData\\Miniconda3\\lib\\site-packages\\sklearn\\metrics\\classification.py:1437: UndefinedMetricWarning: F-score is ill-defined and being set to 0.0 in labels with no predicted samples.\n",
      "  'precision', 'predicted', average, warn_for)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best parameters : {'class_weight': [{1: 0.05}, {0: 0.2873684210526316, 1: 0.7126315789473684}, {0: 0.2873684210526316, 1: 0.7126315789473684}, {0: 0.2873684210526316, 1: 0.7126315789473684}, {0: 0.4555555555555556, 1: 0.5444444444444444}, {0: 0.4555555555555556, 1: 0.5444444444444444}, {0: 0.2873684210526316, 1: 0.7126315789473684}, {0: 0.4555555555555556, 1: 0.5444444444444444}, {0: 0.4555555555555556, 1: 0.5444444444444444}, {0: 0.2873684210526316, 1: 0.7126315789473684}, {1: 0.05}]} score:\n"
     ]
    }
   ],
   "source": [
    "middle = np.linspace(0.1, 0.9,10)\n",
    "minority=np.linspace(0.1, 0.99, 20)\n",
    "majority=np.linspace(0.01, 0.05, 5)\n",
    "\n",
    "gsc = GridSearchCV(\n",
    "    estimator=RandomForestClassifier(n_estimators=100),\n",
    "    param_grid={\n",
    "        'class_weight':\n",
    "#         [{0: x, 1: 1.0-x} for x in weights]\n",
    "        [[{1: z},{0: x, 1:1.0-x},{0: x, 1: 1.0-x}, {0: x, 1: 1.0-x},\n",
    "                                          {0: y, 1: 1.0-y},{0: y, 1: 1.0-y}, {0: x, 1: 1.0-x}, {0: y, 1: 1.0-y},\n",
    "                                          {0: y, 1: 1.0-y}, {0: x, 1: 1.0-x}, {1: z}] \n",
    "                                          for x,y,z in zip(minority, middle, majority)\n",
    "                                         ]\n",
    "    },\n",
    "#     scoring='f1'<--was problem before\n",
    "    scoring='f1_macro'\n",
    "#     scoring='recall_macro'\n",
    "   \n",
    ")\n",
    "grid_result = gsc.fit(X_train, y_train)\n",
    "\n",
    "print(\"Best parameters : %s\" % grid_result.best_params_,'score:')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Random Forest: \n",
      "               precision    recall  f1-score   support\n",
      "\n",
      "           0       1.00      1.00      1.00       643\n",
      "           1       1.00      0.06      0.12        16\n",
      "           2       0.50      0.08      0.14        25\n",
      "           3       0.74      0.67      0.70       150\n",
      "           4       0.45      0.14      0.22        35\n",
      "           5       0.77      0.58      0.66       141\n",
      "           6       0.60      0.23      0.33        40\n",
      "           7       0.71      0.34      0.46        80\n",
      "           8       0.71      0.46      0.56        92\n",
      "           9       1.00      0.38      0.56        13\n",
      "          10       1.00      1.00      1.00       643\n",
      "\n",
      "   micro avg       0.94      0.83      0.88      1878\n",
      "   macro avg       0.77      0.45      0.52      1878\n",
      "weighted avg       0.91      0.83      0.85      1878\n",
      " samples avg       0.96      0.88      0.91      1878\n",
      " \n",
      "\n",
      "Validation score: \n",
      " 0.6034214618973561\n",
      "Minimal Probability difference: 0.5038718291054739\n"
     ]
    }
   ],
   "source": [
    "rf_opt = RandomForestClassifier(n_estimators=150,max_features=132,class_weight=[{1: 0.05}, {0: 0.2873684210526316, 1: 0.7126315789473684}, {0: 0.2873684210526316, 1: 0.7126315789473684}, \n",
    "                                                                                {0: 0.2873684210526316, 1: 0.7126315789473684}, {0: 0.4555555555555556, 1: 0.5444444444444444}, \n",
    "                                                                                {0: 0.4555555555555556, 1: 0.5444444444444444}, {0: 0.2873684210526316, 1: 0.7126315789473684}, \n",
    "                                                                                {0: 0.4555555555555556, 1: 0.5444444444444444}, {0: 0.4555555555555556, 1: 0.5444444444444444}, \n",
    "                                                                                {0: 0.2873684210526316, 1: 0.7126315789473684}, {1: 0.05}]\n",
    "                                         )\n",
    "\n",
    "rf_opt.fit(X_train, y_train)\n",
    "print(\"Random Forest: \\n\", classification_report(y_val, rf_opt.predict(X_val)), \"\\n\")\n",
    "print(\"Validation score: \\n\",rf_opt.score(X_val, y_val))\n",
    "print('Minimal Probability difference:', prob_diff(rf_opt.predict_proba(X_ts3)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "SVM Model Refinement"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "'balanced' class weights to strengthen minority classes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\ProgramData\\Miniconda3\\lib\\site-packages\\sklearn\\multiclass.py:76: UserWarning: Label 0 is present in all training examples.\n",
      "  str(classes[c]))\n",
      "C:\\ProgramData\\Miniconda3\\lib\\site-packages\\sklearn\\multiclass.py:76: UserWarning: Label 10 is present in all training examples.\n",
      "  str(classes[c]))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Support Vector Machine: \n",
      "               precision    recall  f1-score   support\n",
      "\n",
      "           0       1.00      1.00      1.00       643\n",
      "           1       0.31      0.50      0.38        16\n",
      "           2       0.22      0.44      0.29        25\n",
      "           3       0.67      0.73      0.70       150\n",
      "           4       0.29      0.54      0.38        35\n",
      "           5       0.59      0.72      0.65       141\n",
      "           6       0.28      0.53      0.37        40\n",
      "           7       0.48      0.74      0.58        80\n",
      "           8       0.55      0.66      0.60        92\n",
      "           9       0.38      0.62      0.47        13\n",
      "          10       1.00      1.00      1.00       643\n",
      "\n",
      "   micro avg       0.80      0.90      0.85      1878\n",
      "   macro avg       0.52      0.68      0.58      1878\n",
      "weighted avg       0.85      0.90      0.87      1878\n",
      " samples avg       0.87      0.93      0.88      1878\n",
      " \n",
      "\n",
      "Validation score: \n",
      " 0.5225505443234837\n",
      "Minimal Probability difference: 0.9588413367236808\n"
     ]
    }
   ],
   "source": [
    "svm = OneVsRestClassifier(SVC(gamma='auto',probability=True,class_weight='balanced'))\n",
    "svm.fit(X_train, y_train)\n",
    "print(\"Support Vector Machine: \\n\", classification_report(y_val, svm.predict(X_val)), \"\\n\")\n",
    "print(\"Validation score: \\n\",svm.score(X_val, y_val))\n",
    "print('Minimal Probability difference:', prob_diff2(svm.predict_proba(X_ts3)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "GridSearch for kernels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\ProgramData\\Miniconda3\\lib\\site-packages\\sklearn\\multiclass.py:76: UserWarning: Label 0 is present in all training examples.\n",
      "  str(classes[c]))\n",
      "C:\\ProgramData\\Miniconda3\\lib\\site-packages\\sklearn\\multiclass.py:76: UserWarning: Label 10 is present in all training examples.\n",
      "  str(classes[c]))\n",
      "C:\\ProgramData\\Miniconda3\\lib\\site-packages\\sklearn\\multiclass.py:76: UserWarning: Label 0 is present in all training examples.\n",
      "  str(classes[c]))\n",
      "C:\\ProgramData\\Miniconda3\\lib\\site-packages\\sklearn\\multiclass.py:76: UserWarning: Label 10 is present in all training examples.\n",
      "  str(classes[c]))\n",
      "C:\\ProgramData\\Miniconda3\\lib\\site-packages\\sklearn\\multiclass.py:76: UserWarning: Label 0 is present in all training examples.\n",
      "  str(classes[c]))\n",
      "C:\\ProgramData\\Miniconda3\\lib\\site-packages\\sklearn\\multiclass.py:76: UserWarning: Label 10 is present in all training examples.\n",
      "  str(classes[c]))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 1min 54s\n"
     ]
    },
    {
     "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>precision</th>\n",
       "      <th>kernel</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.550783</td>\n",
       "      <td>poly</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.524041</td>\n",
       "      <td>rbf</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.376672</td>\n",
       "      <td>sigmoid</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  precision   kernel\n",
       "0  0.550783     poly\n",
       "1  0.524041      rbf\n",
       "2  0.376672  sigmoid"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "kernels = ['poly', 'rbf', 'sigmoid']\n",
    "scores_SVM = dict()\n",
    "i = 0\n",
    "for k in kernels:    \n",
    "        svm2=OneVsRestClassifier(SVC(gamma='auto',probability=True,class_weight='balanced',kernel=k))\n",
    "        svm2.fit(X_train, y_train)        \n",
    "        scores_SVM[i] = {'precision':precision_score(y_val, svm2.predict(X_val),average='macro'), 'kernel' :k}\n",
    "        i+=1\n",
    "pd.DataFrame(scores_SVM).T"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "KNN Model refinement"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "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>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>10</th>\n",
       "      <th>11</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1.0</td>\n",
       "      <td>0.3750</td>\n",
       "      <td>0.32</td>\n",
       "      <td>0.660000</td>\n",
       "      <td>0.428571</td>\n",
       "      <td>0.687943</td>\n",
       "      <td>0.325</td>\n",
       "      <td>0.5500</td>\n",
       "      <td>0.565217</td>\n",
       "      <td>0.153846</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1.0</td>\n",
       "      <td>0.1250</td>\n",
       "      <td>0.12</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>0.171429</td>\n",
       "      <td>0.411348</td>\n",
       "      <td>0.150</td>\n",
       "      <td>0.3125</td>\n",
       "      <td>0.304348</td>\n",
       "      <td>0.076923</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1.0</td>\n",
       "      <td>0.3125</td>\n",
       "      <td>0.16</td>\n",
       "      <td>0.700000</td>\n",
       "      <td>0.257143</td>\n",
       "      <td>0.645390</td>\n",
       "      <td>0.275</td>\n",
       "      <td>0.4625</td>\n",
       "      <td>0.423913</td>\n",
       "      <td>0.153846</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0625</td>\n",
       "      <td>0.12</td>\n",
       "      <td>0.533333</td>\n",
       "      <td>0.171429</td>\n",
       "      <td>0.489362</td>\n",
       "      <td>0.200</td>\n",
       "      <td>0.3125</td>\n",
       "      <td>0.358696</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0625</td>\n",
       "      <td>0.16</td>\n",
       "      <td>0.693333</td>\n",
       "      <td>0.257143</td>\n",
       "      <td>0.609929</td>\n",
       "      <td>0.225</td>\n",
       "      <td>0.4250</td>\n",
       "      <td>0.445652</td>\n",
       "      <td>0.153846</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     1       2     3         4         5         6      7       8         9  \\\n",
       "1  1.0  0.3750  0.32  0.660000  0.428571  0.687943  0.325  0.5500  0.565217   \n",
       "2  1.0  0.1250  0.12  0.500000  0.171429  0.411348  0.150  0.3125  0.304348   \n",
       "3  1.0  0.3125  0.16  0.700000  0.257143  0.645390  0.275  0.4625  0.423913   \n",
       "4  1.0  0.0625  0.12  0.533333  0.171429  0.489362  0.200  0.3125  0.358696   \n",
       "5  1.0  0.0625  0.16  0.693333  0.257143  0.609929  0.225  0.4250  0.445652   \n",
       "\n",
       "         10   11  \n",
       "1  0.153846  1.0  \n",
       "2  0.076923  1.0  \n",
       "3  0.153846  1.0  \n",
       "4  0.000000  1.0  \n",
       "5  0.153846  1.0  "
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#notebook from class\n",
    "## Calculating Recal scores for multiple \"n-neighbors\"\n",
    "recall_scores = {}\n",
    "for n in [1,2,3,4,5,10,20,50]:\n",
    "     knn2 = KNeighborsClassifier(n_neighbors=n)\n",
    "     knn2.fit(X_train, y_train)\n",
    "     recall_scores[n] = recall_score(y_val, knn2.predict(X_val), average = None)\n",
    "\n",
    "### Put recall scores into DataFrame\n",
    "scores_df = pd.DataFrame(recall_scores).T\n",
    "scores_df\n",
    "scores_df.columns = [str(i) for i in range(1,12)]\n",
    "scores_df.index = scores_df.index.astype(str)\n",
    "scores_df.iloc[:5,:]\n",
    "        \n",
    "# ### Optional plot of recall scores\n",
    "# plt.figure(figsize = (10,10))\n",
    "# for col in scores_df:\n",
    "#      if col != 'n_neighbors':\n",
    "#         plt.plot(scores_df[col], label = col)\n",
    "    \n",
    "# plt.ylabel(\" Recall Score\", fontsize = 12)\n",
    "# plt.xlabel(\"n_neighbors\", fontsize = 12)\n",
    "# plt.legend(title = \"Class Label\",loc='center right',bbox_to_anchor=(1, 0.7));"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model Evaluation (Model Choice)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Summarizing all model results in a dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "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>validation score</th>\n",
       "      <th>macro avg recall</th>\n",
       "      <th>probability difference</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Linear Regression</th>\n",
       "      <td>0.18</td>\n",
       "      <td>0.29</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Support Vector Machine</th>\n",
       "      <td>0.53</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Random Forest base model</th>\n",
       "      <td>0.63</td>\n",
       "      <td>0.46</td>\n",
       "      <td>0.51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>k-nearest Neighbors</th>\n",
       "      <td>0.58</td>\n",
       "      <td>0.46</td>\n",
       "      <td>0.54</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Support Vector Machine balanced</th>\n",
       "      <td>0.52</td>\n",
       "      <td>0.68</td>\n",
       "      <td>0.96</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Random Forest optimized</th>\n",
       "      <td>0.60</td>\n",
       "      <td>0.45</td>\n",
       "      <td>0.50</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                 validation score  macro avg recall  \\\n",
       "Linear Regression                            0.18              0.29   \n",
       "Support Vector Machine                       0.53              0.25   \n",
       "Random Forest base model                     0.63              0.46   \n",
       "k-nearest Neighbors                          0.58              0.46   \n",
       "Support Vector Machine balanced              0.52              0.68   \n",
       "Random Forest optimized                      0.60              0.45   \n",
       "\n",
       "                                 probability difference  \n",
       "Linear Regression                                  0.00  \n",
       "Support Vector Machine                             0.18  \n",
       "Random Forest base model                           0.51  \n",
       "k-nearest Neighbors                                0.54  \n",
       "Support Vector Machine balanced                    0.96  \n",
       "Random Forest optimized                            0.50  "
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "res_indv = {'Linear Regression':[ 0.18,0.29,0], 'Support Vector Machine': [0.53,0.25,0.18],\n",
    "           'Random Forest base model':[ 0.63,0.46,0.51], 'k-nearest Neighbors': [0.58,0.46,0.54],\n",
    "            'Support Vector Machine balanced':[0.52,0.68,0.96], 'Random Forest optimized': [0.6,0.45,0.5]\n",
    "           } \n",
    "\n",
    "results = pd.DataFrame(res_indv,  index=['validation score', 'macro avg recall', \n",
    "                                     'probability difference']) \n",
    "results=results.T\n",
    "results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Making predictions with the model of choice"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions=rf_base.predict(X_ts3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.],\n",
       "       [1., 0., 0., 1., 0., 0., 0., 0., 0., 0., 1.],\n",
       "       [1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.],\n",
       "       [1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.],\n",
       "       [1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.],\n",
       "       [1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.],\n",
       "       [1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.],\n",
       "       [1., 0., 0., 1., 0., 1., 0., 0., 0., 0., 1.],\n",
       "       [1., 0., 0., 1., 0., 0., 0., 0., 0., 0., 1.],\n",
       "       [1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.]])"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictions[:10]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Calculating nonzero samples and total stable compounds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(277, 473)"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nonzero_calc(predictions)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Distribution of class labels predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_pos_pred = [0,1,2,3,4,5,6,7,8,9,10]#np.arange(11)\n",
    "squares_pred = [np.count_nonzero(x) for x in predictions.T]\n",
    "\n",
    "plt.bar(y_pos, squares, align='center', alpha=0.5, color='g')\n",
    "plt.xticks(y_pos)\n",
    "plt.ylabel('# Compund Mixtures')\n",
    "plt.title('Frequency of Stable Compound Mixtures Ratios')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Export to csv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "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>new</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>[1.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>[1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>[1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>[1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>744</th>\n",
       "      <td>[1.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>745</th>\n",
       "      <td>[1.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>746</th>\n",
       "      <td>[1.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>747</th>\n",
       "      <td>[1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>748</th>\n",
       "      <td>[1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>749 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                   new\n",
       "0    [1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...\n",
       "1    [1.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...\n",
       "2    [1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...\n",
       "3    [1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...\n",
       "4    [1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...\n",
       "..                                                 ...\n",
       "744  [1.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, ...\n",
       "745  [1.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...\n",
       "746  [1.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, ...\n",
       "747  [1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...\n",
       "748  [1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...\n",
       "\n",
       "[749 rows x 1 columns]"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_pd= pd.DataFrame(predictions)\n",
    "pred_pd['new']=predictions.tolist()\n",
    "pred_pd=pred_pd.drop(pred_pd.columns[[0,1,2,3,4,5,6,7,8,9,10]],axis=1)\n",
    "pred_pd.to_csv('./pred.csv',index=False)\n",
    "pred_pd"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}