NASA_Engine.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from sklearn.preprocessing import MinMaxScaler"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "names = [\"UnitNo\", \"cycle\", 'Setting1', 'Setting2', 'Setting3','s1','s2','s3','s4','s5', 's6', 's7', 's8', 's9',\n",
    "         's10','s11','s12','s13','s14','s15','s16','s17','s18','s19','s20','s21', 's22','s23', 's24', 's25', 's26']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "train1 = pd.read_csv(\"Data/CMAPSSData/train1.csv\", sep=' ' ,names=names)\n",
    "train2 = pd.read_csv(\"Data/CMAPSSData/train2.csv\", sep=' ' ,names=names)\n",
    "train3 = pd.read_csv(\"Data/CMAPSSData/train3.csv\", sep=' ' ,names=names)\n",
    "train4 = pd.read_csv(\"Data/CMAPSSData/train4.csv\", sep=' ' ,names=names)\n",
    "test1 = pd.read_csv(\"Data/CMAPSSData/test1.csv\", sep=' ' ,names=names)\n",
    "test2 = pd.read_csv(\"Data/CMAPSSData/test2.csv\", sep=' ' ,names=names)\n",
    "test3 = pd.read_csv(\"Data/CMAPSSData/test3.csv\", sep=' ' ,names=names)\n",
    "test4 = pd.read_csv(\"Data/CMAPSSData/test4.csv\", sep=' ' ,names=names)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "rul1 = pd.read_csv(\"Data/CMAPSSData/RUL1.csv\", names=['RUL'])\n",
    "rul2 = pd.read_csv(\"Data/CMAPSSData/RUL2.csv\", names=['RUL'])\n",
    "rul3 = pd.read_csv(\"Data/CMAPSSData/RUL3.csv\", names=['RUL'])\n",
    "rul4 = pd.read_csv(\"Data/CMAPSSData/RUL4.csv\", names=['RUL'])"
   ]
  },
  {
   "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>UnitNo</th>\n",
       "      <th>cycle</th>\n",
       "      <th>Setting1</th>\n",
       "      <th>Setting2</th>\n",
       "      <th>Setting3</th>\n",
       "      <th>s1</th>\n",
       "      <th>s2</th>\n",
       "      <th>s3</th>\n",
       "      <th>s4</th>\n",
       "      <th>s5</th>\n",
       "      <th>...</th>\n",
       "      <th>s17</th>\n",
       "      <th>s18</th>\n",
       "      <th>s19</th>\n",
       "      <th>s20</th>\n",
       "      <th>s21</th>\n",
       "      <th>s22</th>\n",
       "      <th>s23</th>\n",
       "      <th>s24</th>\n",
       "      <th>s25</th>\n",
       "      <th>s26</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>61244</th>\n",
       "      <td>249</td>\n",
       "      <td>251</td>\n",
       "      <td>9.9998</td>\n",
       "      <td>0.2500</td>\n",
       "      <td>100.0</td>\n",
       "      <td>489.05</td>\n",
       "      <td>605.33</td>\n",
       "      <td>1516.36</td>\n",
       "      <td>1315.28</td>\n",
       "      <td>10.52</td>\n",
       "      <td>...</td>\n",
       "      <td>372</td>\n",
       "      <td>2319</td>\n",
       "      <td>100.0</td>\n",
       "      <td>29.11</td>\n",
       "      <td>17.5234</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61245</th>\n",
       "      <td>249</td>\n",
       "      <td>252</td>\n",
       "      <td>0.0028</td>\n",
       "      <td>0.0015</td>\n",
       "      <td>100.0</td>\n",
       "      <td>518.67</td>\n",
       "      <td>643.42</td>\n",
       "      <td>1598.92</td>\n",
       "      <td>1426.77</td>\n",
       "      <td>14.62</td>\n",
       "      <td>...</td>\n",
       "      <td>396</td>\n",
       "      <td>2388</td>\n",
       "      <td>100.0</td>\n",
       "      <td>39.38</td>\n",
       "      <td>23.7151</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61246</th>\n",
       "      <td>249</td>\n",
       "      <td>253</td>\n",
       "      <td>0.0029</td>\n",
       "      <td>0.0000</td>\n",
       "      <td>100.0</td>\n",
       "      <td>518.67</td>\n",
       "      <td>643.68</td>\n",
       "      <td>1607.72</td>\n",
       "      <td>1430.56</td>\n",
       "      <td>14.62</td>\n",
       "      <td>...</td>\n",
       "      <td>395</td>\n",
       "      <td>2388</td>\n",
       "      <td>100.0</td>\n",
       "      <td>39.78</td>\n",
       "      <td>23.8270</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61247</th>\n",
       "      <td>249</td>\n",
       "      <td>254</td>\n",
       "      <td>35.0046</td>\n",
       "      <td>0.8400</td>\n",
       "      <td>100.0</td>\n",
       "      <td>449.44</td>\n",
       "      <td>555.77</td>\n",
       "      <td>1381.29</td>\n",
       "      <td>1148.18</td>\n",
       "      <td>5.48</td>\n",
       "      <td>...</td>\n",
       "      <td>337</td>\n",
       "      <td>2223</td>\n",
       "      <td>100.0</td>\n",
       "      <td>15.26</td>\n",
       "      <td>9.0774</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61248</th>\n",
       "      <td>249</td>\n",
       "      <td>255</td>\n",
       "      <td>42.0030</td>\n",
       "      <td>0.8400</td>\n",
       "      <td>100.0</td>\n",
       "      <td>445.00</td>\n",
       "      <td>549.85</td>\n",
       "      <td>1369.75</td>\n",
       "      <td>1147.45</td>\n",
       "      <td>3.91</td>\n",
       "      <td>...</td>\n",
       "      <td>333</td>\n",
       "      <td>2212</td>\n",
       "      <td>100.0</td>\n",
       "      <td>10.66</td>\n",
       "      <td>6.4341</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 31 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       UnitNo  cycle  Setting1  Setting2  Setting3      s1      s2       s3  \\\n",
       "61244     249    251    9.9998    0.2500     100.0  489.05  605.33  1516.36   \n",
       "61245     249    252    0.0028    0.0015     100.0  518.67  643.42  1598.92   \n",
       "61246     249    253    0.0029    0.0000     100.0  518.67  643.68  1607.72   \n",
       "61247     249    254   35.0046    0.8400     100.0  449.44  555.77  1381.29   \n",
       "61248     249    255   42.0030    0.8400     100.0  445.00  549.85  1369.75   \n",
       "\n",
       "            s4     s5 ...   s17   s18    s19    s20      s21  s22  s23  s24  \\\n",
       "61244  1315.28  10.52 ...   372  2319  100.0  29.11  17.5234  NaN  NaN  NaN   \n",
       "61245  1426.77  14.62 ...   396  2388  100.0  39.38  23.7151  NaN  NaN  NaN   \n",
       "61246  1430.56  14.62 ...   395  2388  100.0  39.78  23.8270  NaN  NaN  NaN   \n",
       "61247  1148.18   5.48 ...   337  2223  100.0  15.26   9.0774  NaN  NaN  NaN   \n",
       "61248  1147.45   3.91 ...   333  2212  100.0  10.66   6.4341  NaN  NaN  NaN   \n",
       "\n",
       "       s25  s26  \n",
       "61244  NaN  NaN  \n",
       "61245  NaN  NaN  \n",
       "61246  NaN  NaN  \n",
       "61247  NaN  NaN  \n",
       "61248  NaN  NaN  \n",
       "\n",
       "[5 rows x 31 columns]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train4.tail()"
   ]
  },
  {
   "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>RUL</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>112</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>98</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>69</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>82</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>91</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   RUL\n",
       "0  112\n",
       "1   98\n",
       "2   69\n",
       "3   82\n",
       "4   91"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rul1.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "260"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train2.UnitNo.nunique()"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(20631, 31)\n",
      "(100, 1)\n",
      "(53759, 31)\n",
      "(259, 1)\n",
      "(24720, 31)\n",
      "(100, 1)\n",
      "(61249, 31)\n",
      "(248, 1)\n"
     ]
    }
   ],
   "source": [
    "print(train1.shape)\n",
    "print(rul1.shape)\n",
    "print(train2.shape)\n",
    "print(rul2.shape)\n",
    "print(train3.shape)\n",
    "print(rul3.shape)\n",
    "print(train4.shape)\n",
    "print(rul4.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(13096, 31)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test1.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# train = pd.concat([train1, train2, train3, train4])\n",
    "# test = pd.concat([test1, test2, test3, test4])\n",
    "# truth_df = pd.concat([rul1,rul2,rul3,rul4])\n",
    "train = train2\n",
    "test = test2\n",
    "truth_df = rul2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "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>UnitNo</th>\n",
       "      <th>cycle</th>\n",
       "      <th>Setting1</th>\n",
       "      <th>Setting2</th>\n",
       "      <th>Setting3</th>\n",
       "      <th>s1</th>\n",
       "      <th>s2</th>\n",
       "      <th>s3</th>\n",
       "      <th>s4</th>\n",
       "      <th>s5</th>\n",
       "      <th>...</th>\n",
       "      <th>s17</th>\n",
       "      <th>s18</th>\n",
       "      <th>s19</th>\n",
       "      <th>s20</th>\n",
       "      <th>s21</th>\n",
       "      <th>s22</th>\n",
       "      <th>s23</th>\n",
       "      <th>s24</th>\n",
       "      <th>s25</th>\n",
       "      <th>s26</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>34.9983</td>\n",
       "      <td>0.8400</td>\n",
       "      <td>100.0</td>\n",
       "      <td>449.44</td>\n",
       "      <td>555.32</td>\n",
       "      <td>1358.61</td>\n",
       "      <td>1137.23</td>\n",
       "      <td>5.48</td>\n",
       "      <td>...</td>\n",
       "      <td>334</td>\n",
       "      <td>2223</td>\n",
       "      <td>100.00</td>\n",
       "      <td>14.73</td>\n",
       "      <td>8.8071</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>41.9982</td>\n",
       "      <td>0.8408</td>\n",
       "      <td>100.0</td>\n",
       "      <td>445.00</td>\n",
       "      <td>549.90</td>\n",
       "      <td>1353.22</td>\n",
       "      <td>1125.78</td>\n",
       "      <td>3.91</td>\n",
       "      <td>...</td>\n",
       "      <td>330</td>\n",
       "      <td>2212</td>\n",
       "      <td>100.00</td>\n",
       "      <td>10.41</td>\n",
       "      <td>6.2665</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>24.9988</td>\n",
       "      <td>0.6218</td>\n",
       "      <td>60.0</td>\n",
       "      <td>462.54</td>\n",
       "      <td>537.31</td>\n",
       "      <td>1256.76</td>\n",
       "      <td>1047.45</td>\n",
       "      <td>7.05</td>\n",
       "      <td>...</td>\n",
       "      <td>309</td>\n",
       "      <td>1915</td>\n",
       "      <td>84.93</td>\n",
       "      <td>14.08</td>\n",
       "      <td>8.6723</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>42.0077</td>\n",
       "      <td>0.8416</td>\n",
       "      <td>100.0</td>\n",
       "      <td>445.00</td>\n",
       "      <td>549.51</td>\n",
       "      <td>1354.03</td>\n",
       "      <td>1126.38</td>\n",
       "      <td>3.91</td>\n",
       "      <td>...</td>\n",
       "      <td>329</td>\n",
       "      <td>2212</td>\n",
       "      <td>100.00</td>\n",
       "      <td>10.59</td>\n",
       "      <td>6.4701</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>25.0005</td>\n",
       "      <td>0.6203</td>\n",
       "      <td>60.0</td>\n",
       "      <td>462.54</td>\n",
       "      <td>537.07</td>\n",
       "      <td>1257.71</td>\n",
       "      <td>1047.93</td>\n",
       "      <td>7.05</td>\n",
       "      <td>...</td>\n",
       "      <td>309</td>\n",
       "      <td>1915</td>\n",
       "      <td>84.93</td>\n",
       "      <td>14.13</td>\n",
       "      <td>8.5286</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 31 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   UnitNo  cycle  Setting1  Setting2  Setting3      s1      s2       s3  \\\n",
       "0       1      1   34.9983    0.8400     100.0  449.44  555.32  1358.61   \n",
       "1       1      2   41.9982    0.8408     100.0  445.00  549.90  1353.22   \n",
       "2       1      3   24.9988    0.6218      60.0  462.54  537.31  1256.76   \n",
       "3       1      4   42.0077    0.8416     100.0  445.00  549.51  1354.03   \n",
       "4       1      5   25.0005    0.6203      60.0  462.54  537.07  1257.71   \n",
       "\n",
       "        s4    s5 ...   s17   s18     s19    s20     s21  s22  s23  s24  s25  \\\n",
       "0  1137.23  5.48 ...   334  2223  100.00  14.73  8.8071  NaN  NaN  NaN  NaN   \n",
       "1  1125.78  3.91 ...   330  2212  100.00  10.41  6.2665  NaN  NaN  NaN  NaN   \n",
       "2  1047.45  7.05 ...   309  1915   84.93  14.08  8.6723  NaN  NaN  NaN  NaN   \n",
       "3  1126.38  3.91 ...   329  2212  100.00  10.59  6.4701  NaN  NaN  NaN  NaN   \n",
       "4  1047.93  7.05 ...   309  1915   84.93  14.13  8.5286  NaN  NaN  NaN  NaN   \n",
       "\n",
       "   s26  \n",
       "0  NaN  \n",
       "1  NaN  \n",
       "2  NaN  \n",
       "3  NaN  \n",
       "4  NaN  \n",
       "\n",
       "[5 rows x 31 columns]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "train.drop(columns=['s22', 's23', 's24', 's25', 's26'], inplace=True)\n",
    "test.drop(columns=['s22', 's23', 's24', 's25', 's26'], inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "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>UnitNo</th>\n",
       "      <th>cycle</th>\n",
       "      <th>Setting1</th>\n",
       "      <th>Setting2</th>\n",
       "      <th>Setting3</th>\n",
       "      <th>s1</th>\n",
       "      <th>s2</th>\n",
       "      <th>s3</th>\n",
       "      <th>s4</th>\n",
       "      <th>s5</th>\n",
       "      <th>...</th>\n",
       "      <th>s13</th>\n",
       "      <th>s14</th>\n",
       "      <th>s15</th>\n",
       "      <th>s16</th>\n",
       "      <th>s17</th>\n",
       "      <th>s18</th>\n",
       "      <th>s19</th>\n",
       "      <th>s20</th>\n",
       "      <th>s21</th>\n",
       "      <th>RUL</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>34.9983</td>\n",
       "      <td>0.8400</td>\n",
       "      <td>100.0</td>\n",
       "      <td>449.44</td>\n",
       "      <td>555.32</td>\n",
       "      <td>1358.61</td>\n",
       "      <td>1137.23</td>\n",
       "      <td>5.48</td>\n",
       "      <td>...</td>\n",
       "      <td>2387.72</td>\n",
       "      <td>8048.56</td>\n",
       "      <td>9.3461</td>\n",
       "      <td>0.02</td>\n",
       "      <td>334</td>\n",
       "      <td>2223</td>\n",
       "      <td>100.00</td>\n",
       "      <td>14.73</td>\n",
       "      <td>8.8071</td>\n",
       "      <td>148</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>41.9982</td>\n",
       "      <td>0.8408</td>\n",
       "      <td>100.0</td>\n",
       "      <td>445.00</td>\n",
       "      <td>549.90</td>\n",
       "      <td>1353.22</td>\n",
       "      <td>1125.78</td>\n",
       "      <td>3.91</td>\n",
       "      <td>...</td>\n",
       "      <td>2387.66</td>\n",
       "      <td>8072.30</td>\n",
       "      <td>9.3774</td>\n",
       "      <td>0.02</td>\n",
       "      <td>330</td>\n",
       "      <td>2212</td>\n",
       "      <td>100.00</td>\n",
       "      <td>10.41</td>\n",
       "      <td>6.2665</td>\n",
       "      <td>147</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>24.9988</td>\n",
       "      <td>0.6218</td>\n",
       "      <td>60.0</td>\n",
       "      <td>462.54</td>\n",
       "      <td>537.31</td>\n",
       "      <td>1256.76</td>\n",
       "      <td>1047.45</td>\n",
       "      <td>7.05</td>\n",
       "      <td>...</td>\n",
       "      <td>2028.03</td>\n",
       "      <td>7864.87</td>\n",
       "      <td>10.8941</td>\n",
       "      <td>0.02</td>\n",
       "      <td>309</td>\n",
       "      <td>1915</td>\n",
       "      <td>84.93</td>\n",
       "      <td>14.08</td>\n",
       "      <td>8.6723</td>\n",
       "      <td>146</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>42.0077</td>\n",
       "      <td>0.8416</td>\n",
       "      <td>100.0</td>\n",
       "      <td>445.00</td>\n",
       "      <td>549.51</td>\n",
       "      <td>1354.03</td>\n",
       "      <td>1126.38</td>\n",
       "      <td>3.91</td>\n",
       "      <td>...</td>\n",
       "      <td>2387.61</td>\n",
       "      <td>8068.66</td>\n",
       "      <td>9.3528</td>\n",
       "      <td>0.02</td>\n",
       "      <td>329</td>\n",
       "      <td>2212</td>\n",
       "      <td>100.00</td>\n",
       "      <td>10.59</td>\n",
       "      <td>6.4701</td>\n",
       "      <td>145</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>25.0005</td>\n",
       "      <td>0.6203</td>\n",
       "      <td>60.0</td>\n",
       "      <td>462.54</td>\n",
       "      <td>537.07</td>\n",
       "      <td>1257.71</td>\n",
       "      <td>1047.93</td>\n",
       "      <td>7.05</td>\n",
       "      <td>...</td>\n",
       "      <td>2028.00</td>\n",
       "      <td>7861.23</td>\n",
       "      <td>10.8963</td>\n",
       "      <td>0.02</td>\n",
       "      <td>309</td>\n",
       "      <td>1915</td>\n",
       "      <td>84.93</td>\n",
       "      <td>14.13</td>\n",
       "      <td>8.5286</td>\n",
       "      <td>144</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 27 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   UnitNo  cycle  Setting1  Setting2  Setting3      s1      s2       s3  \\\n",
       "0       1      1   34.9983    0.8400     100.0  449.44  555.32  1358.61   \n",
       "1       1      2   41.9982    0.8408     100.0  445.00  549.90  1353.22   \n",
       "2       1      3   24.9988    0.6218      60.0  462.54  537.31  1256.76   \n",
       "3       1      4   42.0077    0.8416     100.0  445.00  549.51  1354.03   \n",
       "4       1      5   25.0005    0.6203      60.0  462.54  537.07  1257.71   \n",
       "\n",
       "        s4    s5 ...       s13      s14      s15   s16  s17   s18     s19  \\\n",
       "0  1137.23  5.48 ...   2387.72  8048.56   9.3461  0.02  334  2223  100.00   \n",
       "1  1125.78  3.91 ...   2387.66  8072.30   9.3774  0.02  330  2212  100.00   \n",
       "2  1047.45  7.05 ...   2028.03  7864.87  10.8941  0.02  309  1915   84.93   \n",
       "3  1126.38  3.91 ...   2387.61  8068.66   9.3528  0.02  329  2212  100.00   \n",
       "4  1047.93  7.05 ...   2028.00  7861.23  10.8963  0.02  309  1915   84.93   \n",
       "\n",
       "     s20     s21  RUL  \n",
       "0  14.73  8.8071  148  \n",
       "1  10.41  6.2665  147  \n",
       "2  14.08  8.6723  146  \n",
       "3  10.59  6.4701  145  \n",
       "4  14.13  8.5286  144  \n",
       "\n",
       "[5 rows x 27 columns]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Data Labeling - generate column RUL\n",
    "rul_train = pd.DataFrame(train.groupby('UnitNo')['cycle'].max()).reset_index()\n",
    "rul_train.columns = ['UnitNo', 'max']\n",
    "train = pd.merge(train,rul_train, on='UnitNo', how='left')\n",
    "train['RUL'] = train['max'] - train['cycle']\n",
    "train.drop('max', axis=1, inplace=True)\n",
    "train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(260, 2)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rul_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(259, 1)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "truth_df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "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>UnitNo</th>\n",
       "      <th>cycle</th>\n",
       "      <th>Setting1</th>\n",
       "      <th>Setting2</th>\n",
       "      <th>Setting3</th>\n",
       "      <th>s1</th>\n",
       "      <th>s2</th>\n",
       "      <th>s3</th>\n",
       "      <th>s4</th>\n",
       "      <th>s5</th>\n",
       "      <th>...</th>\n",
       "      <th>s15</th>\n",
       "      <th>s16</th>\n",
       "      <th>s17</th>\n",
       "      <th>s18</th>\n",
       "      <th>s19</th>\n",
       "      <th>s20</th>\n",
       "      <th>s21</th>\n",
       "      <th>RUL</th>\n",
       "      <th>label1</th>\n",
       "      <th>label2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>34.9983</td>\n",
       "      <td>0.8400</td>\n",
       "      <td>100.0</td>\n",
       "      <td>449.44</td>\n",
       "      <td>555.32</td>\n",
       "      <td>1358.61</td>\n",
       "      <td>1137.23</td>\n",
       "      <td>5.48</td>\n",
       "      <td>...</td>\n",
       "      <td>9.3461</td>\n",
       "      <td>0.02</td>\n",
       "      <td>334</td>\n",
       "      <td>2223</td>\n",
       "      <td>100.00</td>\n",
       "      <td>14.73</td>\n",
       "      <td>8.8071</td>\n",
       "      <td>148</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>41.9982</td>\n",
       "      <td>0.8408</td>\n",
       "      <td>100.0</td>\n",
       "      <td>445.00</td>\n",
       "      <td>549.90</td>\n",
       "      <td>1353.22</td>\n",
       "      <td>1125.78</td>\n",
       "      <td>3.91</td>\n",
       "      <td>...</td>\n",
       "      <td>9.3774</td>\n",
       "      <td>0.02</td>\n",
       "      <td>330</td>\n",
       "      <td>2212</td>\n",
       "      <td>100.00</td>\n",
       "      <td>10.41</td>\n",
       "      <td>6.2665</td>\n",
       "      <td>147</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>24.9988</td>\n",
       "      <td>0.6218</td>\n",
       "      <td>60.0</td>\n",
       "      <td>462.54</td>\n",
       "      <td>537.31</td>\n",
       "      <td>1256.76</td>\n",
       "      <td>1047.45</td>\n",
       "      <td>7.05</td>\n",
       "      <td>...</td>\n",
       "      <td>10.8941</td>\n",
       "      <td>0.02</td>\n",
       "      <td>309</td>\n",
       "      <td>1915</td>\n",
       "      <td>84.93</td>\n",
       "      <td>14.08</td>\n",
       "      <td>8.6723</td>\n",
       "      <td>146</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>42.0077</td>\n",
       "      <td>0.8416</td>\n",
       "      <td>100.0</td>\n",
       "      <td>445.00</td>\n",
       "      <td>549.51</td>\n",
       "      <td>1354.03</td>\n",
       "      <td>1126.38</td>\n",
       "      <td>3.91</td>\n",
       "      <td>...</td>\n",
       "      <td>9.3528</td>\n",
       "      <td>0.02</td>\n",
       "      <td>329</td>\n",
       "      <td>2212</td>\n",
       "      <td>100.00</td>\n",
       "      <td>10.59</td>\n",
       "      <td>6.4701</td>\n",
       "      <td>145</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>25.0005</td>\n",
       "      <td>0.6203</td>\n",
       "      <td>60.0</td>\n",
       "      <td>462.54</td>\n",
       "      <td>537.07</td>\n",
       "      <td>1257.71</td>\n",
       "      <td>1047.93</td>\n",
       "      <td>7.05</td>\n",
       "      <td>...</td>\n",
       "      <td>10.8963</td>\n",
       "      <td>0.02</td>\n",
       "      <td>309</td>\n",
       "      <td>1915</td>\n",
       "      <td>84.93</td>\n",
       "      <td>14.13</td>\n",
       "      <td>8.5286</td>\n",
       "      <td>144</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 29 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   UnitNo  cycle  Setting1  Setting2  Setting3      s1      s2       s3  \\\n",
       "0       1      1   34.9983    0.8400     100.0  449.44  555.32  1358.61   \n",
       "1       1      2   41.9982    0.8408     100.0  445.00  549.90  1353.22   \n",
       "2       1      3   24.9988    0.6218      60.0  462.54  537.31  1256.76   \n",
       "3       1      4   42.0077    0.8416     100.0  445.00  549.51  1354.03   \n",
       "4       1      5   25.0005    0.6203      60.0  462.54  537.07  1257.71   \n",
       "\n",
       "        s4    s5   ...        s15   s16  s17   s18     s19    s20     s21  \\\n",
       "0  1137.23  5.48   ...     9.3461  0.02  334  2223  100.00  14.73  8.8071   \n",
       "1  1125.78  3.91   ...     9.3774  0.02  330  2212  100.00  10.41  6.2665   \n",
       "2  1047.45  7.05   ...    10.8941  0.02  309  1915   84.93  14.08  8.6723   \n",
       "3  1126.38  3.91   ...     9.3528  0.02  329  2212  100.00  10.59  6.4701   \n",
       "4  1047.93  7.05   ...    10.8963  0.02  309  1915   84.93  14.13  8.5286   \n",
       "\n",
       "   RUL  label1  label2  \n",
       "0  148       0       0  \n",
       "1  147       0       0  \n",
       "2  146       0       0  \n",
       "3  145       0       0  \n",
       "4  144       0       0  \n",
       "\n",
       "[5 rows x 29 columns]"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# generate label columns for training data\n",
    "w1 = 30\n",
    "w0 = 15\n",
    "\n",
    "# Label1 indicates a failure will occur within the next 30 cycles.\n",
    "# 1 indicates failure, 0 indicates healthy \n",
    "train['label1'] = np.where(train['RUL'] <= w1, 1, 0 )\n",
    "\n",
    "# label2 is multiclass, value 1 is identical to label1,\n",
    "# value 2 indicates failure within 15 cycles\n",
    "train['label2'] = train['label1']\n",
    "train.loc[train['RUL'] <= w0, 'label2'] = 2\n",
    "train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "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>UnitNo</th>\n",
       "      <th>cycle</th>\n",
       "      <th>Setting1</th>\n",
       "      <th>Setting2</th>\n",
       "      <th>Setting3</th>\n",
       "      <th>s1</th>\n",
       "      <th>s2</th>\n",
       "      <th>s3</th>\n",
       "      <th>s4</th>\n",
       "      <th>s5</th>\n",
       "      <th>...</th>\n",
       "      <th>s16</th>\n",
       "      <th>s17</th>\n",
       "      <th>s18</th>\n",
       "      <th>s19</th>\n",
       "      <th>s20</th>\n",
       "      <th>s21</th>\n",
       "      <th>RUL</th>\n",
       "      <th>label1</th>\n",
       "      <th>label2</th>\n",
       "      <th>Unit_norm</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0.833134</td>\n",
       "      <td>0.997625</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.060269</td>\n",
       "      <td>0.181576</td>\n",
       "      <td>0.311201</td>\n",
       "      <td>0.273095</td>\n",
       "      <td>0.146592</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.322917</td>\n",
       "      <td>0.651163</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.156036</td>\n",
       "      <td>0.159082</td>\n",
       "      <td>148</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>0.999767</td>\n",
       "      <td>0.998575</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.131847</td>\n",
       "      <td>0.296600</td>\n",
       "      <td>0.245535</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.281250</td>\n",
       "      <td>0.627907</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.007888</td>\n",
       "      <td>0.014562</td>\n",
       "      <td>147</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>0.595096</td>\n",
       "      <td>0.738480</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.238089</td>\n",
       "      <td>0.016332</td>\n",
       "      <td>0.035297</td>\n",
       "      <td>0.056997</td>\n",
       "      <td>0.293184</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.062500</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.133745</td>\n",
       "      <td>0.151414</td>\n",
       "      <td>146</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>0.999993</td>\n",
       "      <td>0.999525</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.128269</td>\n",
       "      <td>0.298795</td>\n",
       "      <td>0.246979</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.270833</td>\n",
       "      <td>0.627907</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.014060</td>\n",
       "      <td>0.026144</td>\n",
       "      <td>145</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>0.595137</td>\n",
       "      <td>0.736698</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.238089</td>\n",
       "      <td>0.014130</td>\n",
       "      <td>0.037871</td>\n",
       "      <td>0.058152</td>\n",
       "      <td>0.293184</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.062500</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.135460</td>\n",
       "      <td>0.143240</td>\n",
       "      <td>144</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 30 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   UnitNo  cycle  Setting1  Setting2  Setting3        s1        s2        s3  \\\n",
       "0       1      1  0.833134  0.997625       1.0  0.060269  0.181576  0.311201   \n",
       "1       1      2  0.999767  0.998575       1.0  0.000000  0.131847  0.296600   \n",
       "2       1      3  0.595096  0.738480       0.0  0.238089  0.016332  0.035297   \n",
       "3       1      4  0.999993  0.999525       1.0  0.000000  0.128269  0.298795   \n",
       "4       1      5  0.595137  0.736698       0.0  0.238089  0.014130  0.037871   \n",
       "\n",
       "         s4        s5    ...      s16       s17       s18  s19       s20  \\\n",
       "0  0.273095  0.146592    ...      0.0  0.322917  0.651163  1.0  0.156036   \n",
       "1  0.245535  0.000000    ...      0.0  0.281250  0.627907  1.0  0.007888   \n",
       "2  0.056997  0.293184    ...      0.0  0.062500  0.000000  0.0  0.133745   \n",
       "3  0.246979  0.000000    ...      0.0  0.270833  0.627907  1.0  0.014060   \n",
       "4  0.058152  0.293184    ...      0.0  0.062500  0.000000  0.0  0.135460   \n",
       "\n",
       "        s21  RUL  label1  label2  Unit_norm  \n",
       "0  0.159082  148       0       0        0.0  \n",
       "1  0.014562  147       0       0        0.0  \n",
       "2  0.151414  146       0       0        0.0  \n",
       "3  0.026144  145       0       0        0.0  \n",
       "4  0.143240  144       0       0        0.0  \n",
       "\n",
       "[5 rows x 30 columns]"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# MinMax normalization\n",
    "train['Unit_norm'] = train['UnitNo']\n",
    "cols_normalize = train.columns.difference(['UnitNo', 'cycle', 'RUL', 'label1', 'label2'])\n",
    "scaler = MinMaxScaler()\n",
    "norm_train = pd.DataFrame(scaler.fit_transform(train[cols_normalize]), columns=cols_normalize, index=train.index)\n",
    "\n",
    "join_df = train[train.columns.difference(cols_normalize)].join(norm_train)\n",
    "train = join_df.reindex(columns=train.columns)\n",
    "train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "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>UnitNo</th>\n",
       "      <th>cycle</th>\n",
       "      <th>Setting1</th>\n",
       "      <th>Setting2</th>\n",
       "      <th>Setting3</th>\n",
       "      <th>s1</th>\n",
       "      <th>s2</th>\n",
       "      <th>s3</th>\n",
       "      <th>s4</th>\n",
       "      <th>s5</th>\n",
       "      <th>...</th>\n",
       "      <th>s13</th>\n",
       "      <th>s14</th>\n",
       "      <th>s15</th>\n",
       "      <th>s16</th>\n",
       "      <th>s17</th>\n",
       "      <th>s18</th>\n",
       "      <th>s19</th>\n",
       "      <th>s20</th>\n",
       "      <th>s21</th>\n",
       "      <th>Unit_norm</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0.238019</td>\n",
       "      <td>0.297150</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.597937</td>\n",
       "      <td>0.637673</td>\n",
       "      <td>0.686550</td>\n",
       "      <td>0.676888</td>\n",
       "      <td>0.617180</td>\n",
       "      <td>...</td>\n",
       "      <td>0.993662</td>\n",
       "      <td>0.632503</td>\n",
       "      <td>0.114199</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.687500</td>\n",
       "      <td>0.854123</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.625514</td>\n",
       "      <td>0.633951</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>0.476162</td>\n",
       "      <td>0.831354</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.626985</td>\n",
       "      <td>0.663272</td>\n",
       "      <td>0.643289</td>\n",
       "      <td>0.535166</td>\n",
       "      <td>0.507937</td>\n",
       "      <td>...</td>\n",
       "      <td>0.993496</td>\n",
       "      <td>0.487219</td>\n",
       "      <td>0.331283</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.635417</td>\n",
       "      <td>0.864693</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.483882</td>\n",
       "      <td>0.500205</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>0.833282</td>\n",
       "      <td>0.997625</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.060269</td>\n",
       "      <td>0.187815</td>\n",
       "      <td>0.312475</td>\n",
       "      <td>0.251745</td>\n",
       "      <td>0.146592</td>\n",
       "      <td>...</td>\n",
       "      <td>0.992477</td>\n",
       "      <td>0.487171</td>\n",
       "      <td>0.370350</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.312500</td>\n",
       "      <td>0.651163</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.164609</td>\n",
       "      <td>0.165078</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>0.999967</td>\n",
       "      <td>0.998812</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.134324</td>\n",
       "      <td>0.287038</td>\n",
       "      <td>0.250614</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.992394</td>\n",
       "      <td>0.520160</td>\n",
       "      <td>0.388254</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.302083</td>\n",
       "      <td>0.627907</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.005830</td>\n",
       "      <td>0.023186</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>0.595089</td>\n",
       "      <td>0.737886</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.238089</td>\n",
       "      <td>0.010918</td>\n",
       "      <td>0.025599</td>\n",
       "      <td>0.064796</td>\n",
       "      <td>0.293184</td>\n",
       "      <td>...</td>\n",
       "      <td>0.001213</td>\n",
       "      <td>0.041177</td>\n",
       "      <td>0.927248</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.020833</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.141632</td>\n",
       "      <td>0.145822</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 27 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   UnitNo  cycle  Setting1  Setting2  Setting3        s1        s2        s3  \\\n",
       "0       1      1  0.238019  0.297150       1.0  0.597937  0.637673  0.686550   \n",
       "1       1      2  0.476162  0.831354       1.0  0.626985  0.663272  0.643289   \n",
       "2       1      3  0.833282  0.997625       1.0  0.060269  0.187815  0.312475   \n",
       "3       1      4  0.999967  0.998812       1.0  0.000000  0.134324  0.287038   \n",
       "4       1      5  0.595089  0.737886       0.0  0.238089  0.010918  0.025599   \n",
       "\n",
       "         s4        s5    ...           s13       s14       s15  s16       s17  \\\n",
       "0  0.676888  0.617180    ...      0.993662  0.632503  0.114199  1.0  0.687500   \n",
       "1  0.535166  0.507937    ...      0.993496  0.487219  0.331283  0.0  0.635417   \n",
       "2  0.251745  0.146592    ...      0.992477  0.487171  0.370350  0.0  0.312500   \n",
       "3  0.250614  0.000000    ...      0.992394  0.520160  0.388254  0.0  0.302083   \n",
       "4  0.064796  0.293184    ...      0.001213  0.041177  0.927248  0.0  0.020833   \n",
       "\n",
       "        s18  s19       s20       s21  Unit_norm  \n",
       "0  0.854123  1.0  0.625514  0.633951        0.0  \n",
       "1  0.864693  1.0  0.483882  0.500205        0.0  \n",
       "2  0.651163  1.0  0.164609  0.165078        0.0  \n",
       "3  0.627907  1.0  0.005830  0.023186        0.0  \n",
       "4  0.000000  0.0  0.141632  0.145822        0.0  \n",
       "\n",
       "[5 rows x 27 columns]"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test['Unit_norm'] = test['UnitNo']\n",
    "norm_test_df = pd.DataFrame(scaler.transform(test[cols_normalize]), \n",
    "                            columns=cols_normalize, \n",
    "                            index=test.index)\n",
    "test_join_df = test[test.columns.difference(cols_normalize)].join(norm_test_df)\n",
    "test = test_join_df.reindex(columns = test.columns)\n",
    "test = test.reset_index(drop=True)\n",
    "test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "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>UnitNo</th>\n",
       "      <th>cycle</th>\n",
       "      <th>Setting1</th>\n",
       "      <th>Setting2</th>\n",
       "      <th>Setting3</th>\n",
       "      <th>s1</th>\n",
       "      <th>s2</th>\n",
       "      <th>s3</th>\n",
       "      <th>s4</th>\n",
       "      <th>s5</th>\n",
       "      <th>...</th>\n",
       "      <th>s14</th>\n",
       "      <th>s15</th>\n",
       "      <th>s16</th>\n",
       "      <th>s17</th>\n",
       "      <th>s18</th>\n",
       "      <th>s19</th>\n",
       "      <th>s20</th>\n",
       "      <th>s21</th>\n",
       "      <th>Unit_norm</th>\n",
       "      <th>RUL</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0.238019</td>\n",
       "      <td>0.297150</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.597937</td>\n",
       "      <td>0.637673</td>\n",
       "      <td>0.686550</td>\n",
       "      <td>0.676888</td>\n",
       "      <td>0.617180</td>\n",
       "      <td>...</td>\n",
       "      <td>0.632503</td>\n",
       "      <td>0.114199</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.687500</td>\n",
       "      <td>0.854123</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.625514</td>\n",
       "      <td>0.633951</td>\n",
       "      <td>0.0</td>\n",
       "      <td>275</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>0.476162</td>\n",
       "      <td>0.831354</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.626985</td>\n",
       "      <td>0.663272</td>\n",
       "      <td>0.643289</td>\n",
       "      <td>0.535166</td>\n",
       "      <td>0.507937</td>\n",
       "      <td>...</td>\n",
       "      <td>0.487219</td>\n",
       "      <td>0.331283</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.635417</td>\n",
       "      <td>0.864693</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.483882</td>\n",
       "      <td>0.500205</td>\n",
       "      <td>0.0</td>\n",
       "      <td>274</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>0.833282</td>\n",
       "      <td>0.997625</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.060269</td>\n",
       "      <td>0.187815</td>\n",
       "      <td>0.312475</td>\n",
       "      <td>0.251745</td>\n",
       "      <td>0.146592</td>\n",
       "      <td>...</td>\n",
       "      <td>0.487171</td>\n",
       "      <td>0.370350</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.312500</td>\n",
       "      <td>0.651163</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.164609</td>\n",
       "      <td>0.165078</td>\n",
       "      <td>0.0</td>\n",
       "      <td>273</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>0.999967</td>\n",
       "      <td>0.998812</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.134324</td>\n",
       "      <td>0.287038</td>\n",
       "      <td>0.250614</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.520160</td>\n",
       "      <td>0.388254</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.302083</td>\n",
       "      <td>0.627907</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.005830</td>\n",
       "      <td>0.023186</td>\n",
       "      <td>0.0</td>\n",
       "      <td>272</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>0.595089</td>\n",
       "      <td>0.737886</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.238089</td>\n",
       "      <td>0.010918</td>\n",
       "      <td>0.025599</td>\n",
       "      <td>0.064796</td>\n",
       "      <td>0.293184</td>\n",
       "      <td>...</td>\n",
       "      <td>0.041177</td>\n",
       "      <td>0.927248</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.020833</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.141632</td>\n",
       "      <td>0.145822</td>\n",
       "      <td>0.0</td>\n",
       "      <td>271</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 28 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   UnitNo  cycle  Setting1  Setting2  Setting3        s1        s2        s3  \\\n",
       "0       1      1  0.238019  0.297150       1.0  0.597937  0.637673  0.686550   \n",
       "1       1      2  0.476162  0.831354       1.0  0.626985  0.663272  0.643289   \n",
       "2       1      3  0.833282  0.997625       1.0  0.060269  0.187815  0.312475   \n",
       "3       1      4  0.999967  0.998812       1.0  0.000000  0.134324  0.287038   \n",
       "4       1      5  0.595089  0.737886       0.0  0.238089  0.010918  0.025599   \n",
       "\n",
       "         s4        s5 ...        s14       s15  s16       s17       s18  s19  \\\n",
       "0  0.676888  0.617180 ...   0.632503  0.114199  1.0  0.687500  0.854123  1.0   \n",
       "1  0.535166  0.507937 ...   0.487219  0.331283  0.0  0.635417  0.864693  1.0   \n",
       "2  0.251745  0.146592 ...   0.487171  0.370350  0.0  0.312500  0.651163  1.0   \n",
       "3  0.250614  0.000000 ...   0.520160  0.388254  0.0  0.302083  0.627907  1.0   \n",
       "4  0.064796  0.293184 ...   0.041177  0.927248  0.0  0.020833  0.000000  0.0   \n",
       "\n",
       "        s20       s21  Unit_norm  RUL  \n",
       "0  0.625514  0.633951        0.0  275  \n",
       "1  0.483882  0.500205        0.0  274  \n",
       "2  0.164609  0.165078        0.0  273  \n",
       "3  0.005830  0.023186        0.0  272  \n",
       "4  0.141632  0.145822        0.0  271  \n",
       "\n",
       "[5 rows x 28 columns]"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# generate column max for test data\n",
    "rul_test = pd.DataFrame(test.groupby('UnitNo')['cycle'].max()).reset_index()\n",
    "rul_test.columns = ['UnitNo', 'max']\n",
    "truth_df.columns = ['more']\n",
    "truth_df['UnitNo'] = truth_df.index + 1\n",
    "truth_df['max'] = rul_test['max'] + truth_df['more']\n",
    "truth_df.drop('more', axis=1, inplace=True)\n",
    "\n",
    "# generate RUL for test data\n",
    "test = test.merge(truth_df, on=['UnitNo'], how='left')\n",
    "test['RUL'] = test['max'] - test['cycle']\n",
    "test.drop('max', axis=1, inplace=True)\n",
    "test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "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>UnitNo</th>\n",
       "      <th>cycle</th>\n",
       "      <th>Setting1</th>\n",
       "      <th>Setting2</th>\n",
       "      <th>Setting3</th>\n",
       "      <th>s1</th>\n",
       "      <th>s2</th>\n",
       "      <th>s3</th>\n",
       "      <th>s4</th>\n",
       "      <th>s5</th>\n",
       "      <th>...</th>\n",
       "      <th>s16</th>\n",
       "      <th>s17</th>\n",
       "      <th>s18</th>\n",
       "      <th>s19</th>\n",
       "      <th>s20</th>\n",
       "      <th>s21</th>\n",
       "      <th>Unit_norm</th>\n",
       "      <th>RUL</th>\n",
       "      <th>label1</th>\n",
       "      <th>label2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0.238019</td>\n",
       "      <td>0.297150</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.597937</td>\n",
       "      <td>0.637673</td>\n",
       "      <td>0.686550</td>\n",
       "      <td>0.676888</td>\n",
       "      <td>0.617180</td>\n",
       "      <td>...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.687500</td>\n",
       "      <td>0.854123</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.625514</td>\n",
       "      <td>0.633951</td>\n",
       "      <td>0.0</td>\n",
       "      <td>275</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>0.476162</td>\n",
       "      <td>0.831354</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.626985</td>\n",
       "      <td>0.663272</td>\n",
       "      <td>0.643289</td>\n",
       "      <td>0.535166</td>\n",
       "      <td>0.507937</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.635417</td>\n",
       "      <td>0.864693</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.483882</td>\n",
       "      <td>0.500205</td>\n",
       "      <td>0.0</td>\n",
       "      <td>274</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>0.833282</td>\n",
       "      <td>0.997625</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.060269</td>\n",
       "      <td>0.187815</td>\n",
       "      <td>0.312475</td>\n",
       "      <td>0.251745</td>\n",
       "      <td>0.146592</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.312500</td>\n",
       "      <td>0.651163</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.164609</td>\n",
       "      <td>0.165078</td>\n",
       "      <td>0.0</td>\n",
       "      <td>273</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>0.999967</td>\n",
       "      <td>0.998812</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.134324</td>\n",
       "      <td>0.287038</td>\n",
       "      <td>0.250614</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.302083</td>\n",
       "      <td>0.627907</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.005830</td>\n",
       "      <td>0.023186</td>\n",
       "      <td>0.0</td>\n",
       "      <td>272</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>0.595089</td>\n",
       "      <td>0.737886</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.238089</td>\n",
       "      <td>0.010918</td>\n",
       "      <td>0.025599</td>\n",
       "      <td>0.064796</td>\n",
       "      <td>0.293184</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.020833</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.141632</td>\n",
       "      <td>0.145822</td>\n",
       "      <td>0.0</td>\n",
       "      <td>271</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 30 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   UnitNo  cycle  Setting1  Setting2  Setting3        s1        s2        s3  \\\n",
       "0       1      1  0.238019  0.297150       1.0  0.597937  0.637673  0.686550   \n",
       "1       1      2  0.476162  0.831354       1.0  0.626985  0.663272  0.643289   \n",
       "2       1      3  0.833282  0.997625       1.0  0.060269  0.187815  0.312475   \n",
       "3       1      4  0.999967  0.998812       1.0  0.000000  0.134324  0.287038   \n",
       "4       1      5  0.595089  0.737886       0.0  0.238089  0.010918  0.025599   \n",
       "\n",
       "         s4        s5   ...    s16       s17       s18  s19       s20  \\\n",
       "0  0.676888  0.617180   ...    1.0  0.687500  0.854123  1.0  0.625514   \n",
       "1  0.535166  0.507937   ...    0.0  0.635417  0.864693  1.0  0.483882   \n",
       "2  0.251745  0.146592   ...    0.0  0.312500  0.651163  1.0  0.164609   \n",
       "3  0.250614  0.000000   ...    0.0  0.302083  0.627907  1.0  0.005830   \n",
       "4  0.064796  0.293184   ...    0.0  0.020833  0.000000  0.0  0.141632   \n",
       "\n",
       "        s21  Unit_norm  RUL  label1  label2  \n",
       "0  0.633951        0.0  275       0       0  \n",
       "1  0.500205        0.0  274       0       0  \n",
       "2  0.165078        0.0  273       0       0  \n",
       "3  0.023186        0.0  272       0       0  \n",
       "4  0.145822        0.0  271       0       0  \n",
       "\n",
       "[5 rows x 30 columns]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# generate label columns w0 and w1 for test data\n",
    "test['label1'] = np.where(test['RUL'] <= w1, 1, 0 )\n",
    "test['label2'] = test['label1']\n",
    "test.loc[test['RUL'] <= w0, 'label2'] = 2\n",
    "test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(53759, 30)"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Modelling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# pick a large window size of 50 cycles\n",
    "sequence_length = 50"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# function to reshape features into (samples, time steps, features) \n",
    "def gen_sequence(id_df, seq_length, seq_cols):\n",
    "    \"\"\" Only sequences that meet the window-length are considered, no padding is used. This means for testing\n",
    "    we need to drop those which are below the window-length. An alternative would be to pad sequences so that\n",
    "    we can use shorter ones \"\"\"\n",
    "    data_array = id_df[seq_cols].values\n",
    "    num_elements = data_array.shape[0]\n",
    "    for start, stop in zip(range(0, num_elements-seq_length), range(seq_length, num_elements)):\n",
    "        yield data_array[start:stop, :]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['Setting1', 'Setting2', 'Setting3', 'Unit_norm', 's1', 's2', 's3', 's4', 's5', 's6', 's7', 's8', 's9', 's10', 's11', 's12', 's13', 's14', 's15', 's16', 's17', 's18', 's19', 's20', 's21']\n"
     ]
    }
   ],
   "source": [
    "# pick the feature columns \n",
    "sequence_cols = ['Setting1', 'Setting2', 'Setting3', 'Unit_norm']\n",
    "key_cols = ['UnitNo', 'cycle']\n",
    "label_cols = ['label1', 'label2', 'RUL']\n",
    "\n",
    "input_features = test.columns.values.tolist()\n",
    "sensor_cols = [x for x in input_features if x not in set(key_cols)]\n",
    "sensor_cols = [x for x in sensor_cols if x not in set(label_cols)]\n",
    "sensor_cols = [x for x in sensor_cols if x not in set(sequence_cols)]\n",
    "\n",
    "# The time is sequenced along\n",
    "# This may be a silly way to get these column names, but it's relatively clear\n",
    "sequence_cols.extend(sensor_cols)\n",
    "\n",
    "print(sequence_cols)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(40759, 50, 25)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# generator for the sequences\n",
    "seq_gen = (list(gen_sequence(train[train['UnitNo']==id], sequence_length, sequence_cols)) \n",
    "           for id in train['UnitNo'].unique())\n",
    "\n",
    "# generate sequences and convert to numpy array\n",
    "seq_array = np.concatenate(list(seq_gen)).astype(np.float32)\n",
    "seq_array.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "# function to generate labels\n",
    "def gen_labels(id_df, seq_length, label):\n",
    "    data_array = id_df[label].values\n",
    "    num_elements = data_array.shape[0]\n",
    "    return data_array[seq_length:num_elements, :]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(40759, 1)"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# generate labels\n",
    "label_gen = [gen_labels(train[train['UnitNo']==id], sequence_length, ['label1']) \n",
    "             for id in train['UnitNo'].unique()]\n",
    "label_array = np.concatenate(label_gen).astype(np.float32)\n",
    "label_array.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## LSTM Network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/jaromir/anaconda3/envs/keras/lib/python3.6/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n",
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import keras\n",
    "from sklearn.metrics import confusion_matrix, recall_score, precision_score\n",
    "from keras.models import Sequential\n",
    "from sklearn import datasets\n",
    "from keras.layers import Dense, Dropout, LSTM, Activation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm_11 (LSTM)               (None, 50, 100)           50400     \n",
      "_________________________________________________________________\n",
      "dropout_11 (Dropout)         (None, 50, 100)           0         \n",
      "_________________________________________________________________\n",
      "lstm_12 (LSTM)               (None, 50)                30200     \n",
      "_________________________________________________________________\n",
      "dropout_12 (Dropout)         (None, 50)                0         \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 1)                 51        \n",
      "=================================================================\n",
      "Total params: 80,651\n",
      "Trainable params: 80,651\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "# build the network\n",
    "# Feature weights\n",
    "nb_features = seq_array.shape[2]\n",
    "nb_out = label_array.shape[1]\n",
    "\n",
    "# LSTM model\n",
    "model = Sequential()\n",
    "\n",
    "# The first layer\n",
    "model.add(LSTM(\n",
    "         input_shape=(sequence_length, nb_features),\n",
    "         units=100,\n",
    "         return_sequences=True))\n",
    "\n",
    "# Plus a 20% dropout rate\n",
    "model.add(Dropout(0.2))\n",
    "\n",
    "# The second layer\n",
    "model.add(LSTM(\n",
    "          units=50,\n",
    "          return_sequences=False))\n",
    "\n",
    "# Plus a 20% dropout rate\n",
    "model.add(Dropout(0.2))\n",
    "\n",
    "# Dense sigmoid layer\n",
    "model.add(Dense(units=nb_out, activation='sigmoid'))\n",
    "\n",
    "# With adam optimizer and a binary crossentropy loss. We will opimize for model accuracy.\n",
    "model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "# Verify the architecture \n",
    "print(model.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 36683 samples, validate on 4076 samples\n",
      "Epoch 1/20\n",
      "36683/36683 [==============================] - 22s 605us/step - loss: 0.4334 - acc: 0.8201 - val_loss: 0.2600 - val_acc: 0.9041\n",
      "Epoch 2/20\n",
      "36683/36683 [==============================] - 22s 587us/step - loss: 0.2763 - acc: 0.8906 - val_loss: 0.2329 - val_acc: 0.9051\n",
      "Epoch 3/20\n",
      "36683/36683 [==============================] - 22s 596us/step - loss: 0.2492 - acc: 0.8993 - val_loss: 0.2013 - val_acc: 0.9178\n",
      "Epoch 4/20\n",
      "36683/36683 [==============================] - 22s 601us/step - loss: 0.2377 - acc: 0.9035 - val_loss: 0.1773 - val_acc: 0.9242\n",
      "Epoch 5/20\n",
      "36683/36683 [==============================] - 22s 598us/step - loss: 0.2146 - acc: 0.9105 - val_loss: 0.1520 - val_acc: 0.9328\n",
      "Epoch 6/20\n",
      "36683/36683 [==============================] - 21s 585us/step - loss: 0.2263 - acc: 0.9097 - val_loss: 0.2405 - val_acc: 0.9024\n",
      "Epoch 7/20\n",
      "36683/36683 [==============================] - 21s 585us/step - loss: 0.2101 - acc: 0.9134 - val_loss: 0.2535 - val_acc: 0.9048\n",
      "Epoch 8/20\n",
      "36683/36683 [==============================] - 22s 588us/step - loss: 0.1976 - acc: 0.9174 - val_loss: 0.1619 - val_acc: 0.9335\n",
      "Epoch 9/20\n",
      "36683/36683 [==============================] - 22s 592us/step - loss: 0.1666 - acc: 0.9312 - val_loss: 0.2113 - val_acc: 0.9134\n",
      "Epoch 10/20\n",
      "36683/36683 [==============================] - 22s 598us/step - loss: 0.1418 - acc: 0.9409 - val_loss: 0.1857 - val_acc: 0.9254\n",
      "Epoch 11/20\n",
      "36683/36683 [==============================] - 22s 592us/step - loss: 0.1405 - acc: 0.9411 - val_loss: 0.1336 - val_acc: 0.9473\n",
      "Epoch 12/20\n",
      "36683/36683 [==============================] - 22s 597us/step - loss: 0.1291 - acc: 0.9454 - val_loss: 0.1724 - val_acc: 0.9374\n",
      "Epoch 13/20\n",
      "36683/36683 [==============================] - 22s 587us/step - loss: 0.1697 - acc: 0.9288 - val_loss: 0.1410 - val_acc: 0.9379\n",
      "Epoch 14/20\n",
      "36683/36683 [==============================] - 22s 595us/step - loss: 0.1246 - acc: 0.9483 - val_loss: 0.1129 - val_acc: 0.9475\n",
      "Epoch 15/20\n",
      "36683/36683 [==============================] - 22s 597us/step - loss: 0.1297 - acc: 0.9456 - val_loss: 0.1223 - val_acc: 0.9529\n",
      "Epoch 16/20\n",
      "36683/36683 [==============================] - 22s 596us/step - loss: 0.1172 - acc: 0.9511 - val_loss: 0.1154 - val_acc: 0.9490\n",
      "Epoch 17/20\n",
      "36683/36683 [==============================] - 22s 598us/step - loss: 0.1108 - acc: 0.9540 - val_loss: 0.1308 - val_acc: 0.9431\n",
      "Epoch 18/20\n",
      "36683/36683 [==============================] - 22s 598us/step - loss: 0.1185 - acc: 0.9502 - val_loss: 0.1154 - val_acc: 0.9502\n",
      "Epoch 19/20\n",
      "36683/36683 [==============================] - 22s 598us/step - loss: 0.1088 - acc: 0.9548 - val_loss: 0.1229 - val_acc: 0.9497\n",
      "Epoch 20/20\n",
      "36683/36683 [==============================] - 22s 601us/step - loss: 0.1077 - acc: 0.9548 - val_loss: 0.1194 - val_acc: 0.9455\n",
      "CPU times: user 51min 37s, sys: 2min 27s, total: 54min 5s\n",
      "Wall time: 7min 17s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "# fit the network\n",
    "history = model.fit(seq_array, # Training features\n",
    "          label_array, # Training labels\n",
    "          epochs=20,   # We'll stop after 10 epochs\n",
    "          batch_size=200, # \n",
    "          validation_split=0.1, # Use 10% of data to evaluate the loss. (val_loss)\n",
    "            verbose=1, #\n",
    "          callbacks = [keras.callbacks.EarlyStopping(monitor='val_loss', # Monitor the validation loss\n",
    "                                                     min_delta=0,    # until it doesn't change (or gets worse)\n",
    "                                                     patience=10,  # patience > 1 so it continutes if it is not consistently improving\n",
    "                                                     verbose=0, \n",
    "                                                     mode='auto')]) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "acc = history.history['acc']\n",
    "val_acc = history.history['val_acc']\n",
    "loss = history.history['loss']\n",
    "val_loss = history.history['val_loss']\n",
    "\n",
    "epochs = range(1, len(acc)+1)\n",
    "\n",
    "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.legend()\n",
    "\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "40759/40759 [==============================] - 8s 191us/step\n",
      "Training Accurracy: 0.9578498004726358\n"
     ]
    }
   ],
   "source": [
    "# training metrics\n",
    "scores = model.evaluate(seq_array, label_array, verbose=1, batch_size=200)\n",
    "print('Training Accurracy: {}'.format(scores[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "40759/40759 [==============================] - 8s 191us/step\n",
      "Training Confusion matrix\n",
      "- x-axis is predicted labels.\n",
      "- y-axis is true labels\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[32281,   418],\n",
       "       [ 1300,  6760]])"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# make predictions and compute confusion matrix\n",
    "y_pred = model.predict_classes(seq_array,verbose=1, batch_size=200)\n",
    "y_true = label_array\n",
    "print('Training Confusion matrix\\n- x-axis is predicted labels.\\n- y-axis is true labels')\n",
    "cm = confusion_matrix(y_true, y_pred)\n",
    "cm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Precision:  0.941766508776818 \n",
      " Training Recall:  0.8387096774193549 \n",
      " Training F1 Score: 0.8872555453471584\n"
     ]
    }
   ],
   "source": [
    "# compute precision and recall\n",
    "precision = precision_score(y_true, y_pred)\n",
    "recall = recall_score(y_true, y_pred)\n",
    "f1 = 2 * (precision * recall) / (precision + recall)\n",
    "print( 'Training Precision: ', precision, '\\n', 'Training Recall: ', recall, '\\n', 'Training F1 Score:', f1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "# LSTM model\n",
    "model2 = Sequential()\n",
    "\n",
    "# The first layer\n",
    "model2.add(LSTM(\n",
    "         input_shape=(sequence_length, nb_features),\n",
    "         units=100,\n",
    "         return_sequences=True))\n",
    "\n",
    "# Plus a 20% dropout rate\n",
    "model2.add(Dropout(0.2))\n",
    "\n",
    "# The second layer\n",
    "model2.add(LSTM(\n",
    "          units=50,\n",
    "          return_sequences=False))\n",
    "\n",
    "# Plus a 20% dropout rate\n",
    "model2.add(Dropout(0.2))\n",
    "\n",
    "# Dense sigmoid layer\n",
    "model2.add(Dense(units=nb_out, activation='sigmoid'))\n",
    "\n",
    "# With adam optimizer and a binary crossentropy loss. We will opimize for model accuracy.\n",
    "model2.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 36683 samples, validate on 4076 samples\n",
      "Epoch 1/30\n",
      "36683/36683 [==============================] - 20s 557us/step - loss: 0.4434 - acc: 0.8176 - val_loss: 0.2410 - val_acc: 0.8970\n",
      "Epoch 2/30\n",
      "36683/36683 [==============================] - 20s 542us/step - loss: 0.2879 - acc: 0.8864 - val_loss: 0.1968 - val_acc: 0.9171\n",
      "Epoch 3/30\n",
      "36683/36683 [==============================] - 20s 540us/step - loss: 0.2414 - acc: 0.9029 - val_loss: 0.1900 - val_acc: 0.9217\n",
      "Epoch 4/30\n",
      "36683/36683 [==============================] - 20s 541us/step - loss: 0.2202 - acc: 0.9087 - val_loss: 0.1597 - val_acc: 0.9279\n",
      "Epoch 5/30\n",
      "36683/36683 [==============================] - 20s 540us/step - loss: 0.2082 - acc: 0.9133 - val_loss: 0.1832 - val_acc: 0.9217\n",
      "Epoch 6/30\n",
      "36683/36683 [==============================] - 20s 542us/step - loss: 0.2036 - acc: 0.9149 - val_loss: 0.2644 - val_acc: 0.8977\n",
      "Epoch 7/30\n",
      "36683/36683 [==============================] - 20s 545us/step - loss: 0.2393 - acc: 0.9022 - val_loss: 0.1570 - val_acc: 0.9262\n",
      "Epoch 8/30\n",
      "36683/36683 [==============================] - 20s 542us/step - loss: 0.1934 - acc: 0.9193 - val_loss: 0.1730 - val_acc: 0.9215\n",
      "Epoch 9/30\n",
      "36683/36683 [==============================] - 20s 543us/step - loss: 0.1645 - acc: 0.9320 - val_loss: 0.1485 - val_acc: 0.9320\n",
      "Epoch 10/30\n",
      "36683/36683 [==============================] - 20s 544us/step - loss: 0.1616 - acc: 0.9336 - val_loss: 0.1459 - val_acc: 0.9352\n",
      "Epoch 11/30\n",
      "36683/36683 [==============================] - 20s 543us/step - loss: 0.1626 - acc: 0.9339 - val_loss: 0.1537 - val_acc: 0.9308\n",
      "Epoch 12/30\n",
      "36683/36683 [==============================] - 20s 543us/step - loss: 0.1394 - acc: 0.9436 - val_loss: 0.1276 - val_acc: 0.9423\n",
      "Epoch 13/30\n",
      "36683/36683 [==============================] - 20s 544us/step - loss: 0.1203 - acc: 0.9502 - val_loss: 0.2562 - val_acc: 0.8903\n",
      "Epoch 14/30\n",
      "36683/36683 [==============================] - 20s 542us/step - loss: 0.1300 - acc: 0.9456 - val_loss: 0.1420 - val_acc: 0.9379\n",
      "Epoch 15/30\n",
      "36683/36683 [==============================] - 20s 546us/step - loss: 0.1203 - acc: 0.9499 - val_loss: 0.1207 - val_acc: 0.9468\n",
      "Epoch 16/30\n",
      "36683/36683 [==============================] - 20s 544us/step - loss: 0.1107 - acc: 0.9531 - val_loss: 0.1320 - val_acc: 0.9404\n",
      "Epoch 17/30\n",
      "36683/36683 [==============================] - 20s 540us/step - loss: 0.1071 - acc: 0.9532 - val_loss: 0.1805 - val_acc: 0.9208\n",
      "Epoch 18/30\n",
      "36683/36683 [==============================] - 20s 551us/step - loss: 0.1020 - acc: 0.9573 - val_loss: 0.1443 - val_acc: 0.9443\n",
      "Epoch 19/30\n",
      "36683/36683 [==============================] - 20s 543us/step - loss: 0.1135 - acc: 0.9521 - val_loss: 0.1707 - val_acc: 0.9271\n",
      "Epoch 20/30\n",
      "36683/36683 [==============================] - 20s 549us/step - loss: 0.1184 - acc: 0.9517 - val_loss: 0.1376 - val_acc: 0.9404\n",
      "CPU times: user 10min 50s, sys: 49.2 s, total: 11min 40s\n",
      "Wall time: 6min 40s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "# fit the network\n",
    "history2 = model2.fit(seq_array, # Training features\n",
    "          label_array, # Training labels\n",
    "          epochs=30,   # We'll stop after 10 epochs\n",
    "          batch_size=200, # \n",
    "          validation_split=0.10, # Use 10% of data to evaluate the loss. (val_loss)\n",
    "          verbose=1, #\n",
    "          callbacks = [keras.callbacks.EarlyStopping(monitor='val_loss', # Monitor the validation loss\n",
    "                                                     min_delta=0,    # until it doesn't change (or gets worse)\n",
    "                                                     patience=5,  # patience > 1 so it continutes if it is not consistently improving\n",
    "                                                     verbose=0, \n",
    "                                                     mode='auto')]) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "acc = history2.history['acc']\n",
    "val_acc = history2.history['val_acc']\n",
    "loss = history2.history['loss']\n",
    "val_loss = history2.history['val_loss']\n",
    "\n",
    "epochs = range(1, len(acc)+1)\n",
    "\n",
    "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.legend()\n",
    "\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "40759/40759 [==============================] - 6s 139us/step\n",
      "Training Accurracy: 0.9643269010125258\n"
     ]
    }
   ],
   "source": [
    "# training metrics\n",
    "scores = model.evaluate(seq_array, label_array, verbose=1, batch_size=200)\n",
    "print('Training Accurracy: {}'.format(scores[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "40759/40759 [==============================] - 6s 136us/step\n",
      "Training Confusion matrix\n",
      "- x-axis is predicted labels.\n",
      "- y-axis is true labels\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[32294,   405],\n",
       "       [ 1049,  7011]])"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# make predictions and compute confusion matrix\n",
    "y_pred = model.predict_classes(seq_array,verbose=1, batch_size=200)\n",
    "y_true = label_array\n",
    "print('Training Confusion matrix\\n- x-axis is predicted labels.\\n- y-axis is true labels')\n",
    "cm = confusion_matrix(y_true, y_pred)\n",
    "cm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Precision:  0.9453883495145631 \n",
      " Training Recall:  0.8698511166253102 \n",
      " Training F1 Score: 0.9060480744378392\n"
     ]
    }
   ],
   "source": [
    "# compute precision and recall\n",
    "precision = precision_score(y_true, y_pred)\n",
    "recall = recall_score(y_true, y_pred)\n",
    "f1 = 2 * (precision * recall) / (precision + recall)\n",
    "print( 'Training Precision: ', precision, '\\n', 'Training Recall: ', recall, '\\n', 'Training F1 Score:', f1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[5.95253285e-01, 7.36342043e-01, 0.00000000e+00, 0.00000000e+00,\n",
       "        2.38088774e-01, 1.82585558e-02, 5.24718949e-02, 7.11259808e-02,\n",
       "        2.93183940e-01, 2.08805031e-01, 9.16424037e-02, 2.55479076e-03,\n",
       "        4.73050971e-02, 2.70270270e-02, 7.24755700e-02, 8.92073557e-02,\n",
       "        3.52743407e-03, 1.40476984e-01, 9.68951377e-01, 0.00000000e+00,\n",
       "        7.29166667e-02, 0.00000000e+00, 0.00000000e+00, 1.39231824e-01,\n",
       "        1.44821270e-01],\n",
       "       [9.99923824e-01, 9.97624703e-01, 1.00000000e+00, 0.00000000e+00,\n",
       "        0.00000000e+00, 1.35150014e-01, 3.01476365e-01, 2.67606990e-01,\n",
       "        0.00000000e+00, 6.28930818e-04, 4.00935516e-03, 6.28795237e-01,\n",
       "        3.12673766e-01, 2.43243243e-01, 5.12214984e-01, 3.24667089e-03,\n",
       "        9.94956872e-01, 6.57923549e-01, 3.91146749e-01, 0.00000000e+00,\n",
       "        3.12500000e-01, 6.27906977e-01, 1.00000000e+00, 6.17283951e-03,\n",
       "        1.59616829e-02],\n",
       "       [2.38230813e-01, 2.97030879e-01, 1.00000000e+00, 0.00000000e+00,\n",
       "        5.97936745e-01, 6.41251491e-01, 7.34850332e-01, 7.25220238e-01,\n",
       "        6.17180205e-01, 6.15723270e-01, 6.12810844e-01, 8.53722393e-01,\n",
       "        6.83627347e-01, 8.91891892e-01, 7.86644951e-01, 6.13189601e-01,\n",
       "        9.94047455e-01, 7.58556672e-01, 1.50080551e-01, 1.00000000e+00,\n",
       "        7.08333333e-01, 8.54122622e-01, 1.00000000e+00, 6.21399177e-01,\n",
       "        6.30440966e-01]])"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test[test['UnitNo']==1][sequence_cols].values[-3:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1\n",
      "258\n",
      "True\n",
      "2\n",
      "55\n",
      "True\n",
      "3\n",
      "165\n",
      "True\n",
      "4\n",
      "86\n",
      "True\n",
      "5\n",
      "148\n",
      "True\n",
      "6\n",
      "63\n",
      "True\n",
      "7\n",
      "184\n",
      "True\n",
      "8\n",
      "173\n",
      "True\n",
      "9\n",
      "161\n",
      "True\n",
      "10\n",
      "84\n",
      "True\n",
      "11\n",
      "162\n",
      "True\n",
      "12\n",
      "74\n",
      "True\n",
      "13\n",
      "164\n",
      "True\n",
      "14\n",
      "160\n",
      "True\n",
      "15\n",
      "165\n",
      "True\n",
      "16\n",
      "88\n",
      "True\n",
      "17\n",
      "264\n",
      "True\n",
      "18\n",
      "40\n",
      "False\n",
      "19\n",
      "172\n",
      "True\n",
      "20\n",
      "101\n",
      "True\n",
      "21\n",
      "128\n",
      "True\n",
      "22\n",
      "159\n",
      "True\n",
      "23\n",
      "49\n",
      "False\n",
      "24\n",
      "210\n",
      "True\n",
      "25\n",
      "147\n",
      "True\n",
      "26\n",
      "76\n",
      "True\n",
      "27\n",
      "68\n",
      "True\n",
      "28\n",
      "23\n",
      "False\n",
      "29\n",
      "101\n",
      "True\n",
      "30\n",
      "184\n",
      "True\n",
      "31\n",
      "123\n",
      "True\n",
      "32\n",
      "170\n",
      "True\n",
      "33\n",
      "135\n",
      "True\n",
      "34\n",
      "52\n",
      "True\n",
      "35\n",
      "77\n",
      "True\n",
      "36\n",
      "163\n",
      "True\n",
      "37\n",
      "49\n",
      "False\n",
      "38\n",
      "127\n",
      "True\n",
      "39\n",
      "149\n",
      "True\n",
      "40\n",
      "183\n",
      "True\n",
      "41\n",
      "210\n",
      "True\n",
      "42\n",
      "137\n",
      "True\n",
      "43\n",
      "184\n",
      "True\n",
      "44\n",
      "164\n",
      "True\n",
      "45\n",
      "195\n",
      "True\n",
      "46\n",
      "132\n",
      "True\n",
      "47\n",
      "81\n",
      "True\n",
      "48\n",
      "132\n",
      "True\n",
      "49\n",
      "65\n",
      "True\n",
      "50\n",
      "164\n",
      "True\n",
      "51\n",
      "37\n",
      "False\n",
      "52\n",
      "165\n",
      "True\n",
      "53\n",
      "149\n",
      "True\n",
      "54\n",
      "74\n",
      "True\n",
      "55\n",
      "123\n",
      "True\n",
      "56\n",
      "73\n",
      "True\n",
      "57\n",
      "76\n",
      "True\n",
      "58\n",
      "247\n",
      "True\n",
      "59\n",
      "22\n",
      "False\n",
      "60\n",
      "144\n",
      "True\n",
      "61\n",
      "70\n",
      "True\n",
      "62\n",
      "205\n",
      "True\n",
      "63\n",
      "174\n",
      "True\n",
      "64\n",
      "233\n",
      "True\n",
      "65\n",
      "367\n",
      "True\n",
      "66\n",
      "63\n",
      "True\n",
      "67\n",
      "146\n",
      "True\n",
      "68\n",
      "218\n",
      "True\n",
      "69\n",
      "158\n",
      "True\n",
      "70\n",
      "264\n",
      "True\n",
      "71\n",
      "65\n",
      "True\n",
      "72\n",
      "143\n",
      "True\n",
      "73\n",
      "151\n",
      "True\n",
      "74\n",
      "103\n",
      "True\n",
      "75\n",
      "192\n",
      "True\n",
      "76\n",
      "222\n",
      "True\n",
      "77\n",
      "67\n",
      "True\n",
      "78\n",
      "128\n",
      "True\n",
      "79\n",
      "113\n",
      "True\n",
      "80\n",
      "278\n",
      "True\n",
      "81\n",
      "76\n",
      "True\n",
      "82\n",
      "96\n",
      "True\n",
      "83\n",
      "21\n",
      "False\n",
      "84\n",
      "204\n",
      "True\n",
      "85\n",
      "183\n",
      "True\n",
      "86\n",
      "83\n",
      "True\n",
      "87\n",
      "131\n",
      "True\n",
      "88\n",
      "166\n",
      "True\n",
      "89\n",
      "59\n",
      "True\n",
      "90\n",
      "163\n",
      "True\n",
      "91\n",
      "278\n",
      "True\n",
      "92\n",
      "123\n",
      "True\n",
      "93\n",
      "76\n",
      "True\n",
      "94\n",
      "117\n",
      "True\n",
      "95\n",
      "78\n",
      "True\n",
      "96\n",
      "140\n",
      "True\n",
      "97\n",
      "213\n",
      "True\n",
      "98\n",
      "113\n",
      "True\n",
      "99\n",
      "163\n",
      "True\n",
      "100\n",
      "149\n",
      "True\n",
      "101\n",
      "38\n",
      "False\n",
      "102\n",
      "122\n",
      "True\n",
      "103\n",
      "186\n",
      "True\n",
      "104\n",
      "141\n",
      "True\n",
      "105\n",
      "61\n",
      "True\n",
      "106\n",
      "72\n",
      "True\n",
      "107\n",
      "132\n",
      "True\n",
      "108\n",
      "174\n",
      "True\n",
      "109\n",
      "144\n",
      "True\n",
      "110\n",
      "219\n",
      "True\n",
      "111\n",
      "131\n",
      "True\n",
      "112\n",
      "109\n",
      "True\n",
      "113\n",
      "45\n",
      "False\n",
      "114\n",
      "153\n",
      "True\n",
      "115\n",
      "210\n",
      "True\n",
      "116\n",
      "171\n",
      "True\n",
      "117\n",
      "55\n",
      "True\n",
      "118\n",
      "97\n",
      "True\n",
      "119\n",
      "202\n",
      "True\n",
      "120\n",
      "160\n",
      "True\n",
      "121\n",
      "59\n",
      "True\n",
      "122\n",
      "29\n",
      "False\n",
      "123\n",
      "61\n",
      "True\n",
      "124\n",
      "149\n",
      "True\n",
      "125\n",
      "79\n",
      "True\n",
      "126\n",
      "99\n",
      "True\n",
      "127\n",
      "126\n",
      "True\n",
      "128\n",
      "110\n",
      "True\n",
      "129\n",
      "218\n",
      "True\n",
      "130\n",
      "141\n",
      "True\n",
      "131\n",
      "73\n",
      "True\n",
      "132\n",
      "122\n",
      "True\n",
      "133\n",
      "186\n",
      "True\n",
      "134\n",
      "186\n",
      "True\n",
      "135\n",
      "120\n",
      "True\n",
      "136\n",
      "77\n",
      "True\n",
      "137\n",
      "226\n",
      "True\n",
      "138\n",
      "200\n",
      "True\n",
      "139\n",
      "116\n",
      "True\n",
      "140\n",
      "306\n",
      "True\n",
      "141\n",
      "69\n",
      "True\n",
      "142\n",
      "163\n",
      "True\n",
      "143\n",
      "226\n",
      "True\n",
      "144\n",
      "143\n",
      "True\n",
      "145\n",
      "142\n",
      "True\n",
      "146\n",
      "144\n",
      "True\n",
      "147\n",
      "78\n",
      "True\n",
      "148\n",
      "157\n",
      "True\n",
      "149\n",
      "70\n",
      "True\n",
      "150\n",
      "196\n",
      "True\n",
      "151\n",
      "141\n",
      "True\n",
      "152\n",
      "90\n",
      "True\n",
      "153\n",
      "146\n",
      "True\n",
      "154\n",
      "58\n",
      "True\n",
      "155\n",
      "72\n",
      "True\n",
      "156\n",
      "75\n",
      "True\n",
      "157\n",
      "161\n",
      "True\n",
      "158\n",
      "171\n",
      "True\n",
      "159\n",
      "187\n",
      "True\n",
      "160\n",
      "124\n",
      "True\n",
      "161\n",
      "174\n",
      "True\n",
      "162\n",
      "185\n",
      "True\n",
      "163\n",
      "144\n",
      "True\n",
      "164\n",
      "199\n",
      "True\n",
      "165\n",
      "35\n",
      "False\n",
      "166\n",
      "97\n",
      "True\n",
      "167\n",
      "78\n",
      "True\n",
      "168\n",
      "107\n",
      "True\n",
      "169\n",
      "80\n",
      "True\n",
      "170\n",
      "85\n",
      "True\n",
      "171\n",
      "243\n",
      "True\n",
      "172\n",
      "128\n",
      "True\n",
      "173\n",
      "117\n",
      "True\n",
      "174\n",
      "51\n",
      "True\n",
      "175\n",
      "83\n",
      "True\n",
      "176\n",
      "200\n",
      "True\n",
      "177\n",
      "191\n",
      "True\n",
      "178\n",
      "54\n",
      "True\n",
      "179\n",
      "156\n",
      "True\n",
      "180\n",
      "154\n",
      "True\n",
      "181\n",
      "199\n",
      "True\n",
      "182\n",
      "134\n",
      "True\n",
      "183\n",
      "150\n",
      "True\n",
      "184\n",
      "64\n",
      "True\n",
      "185\n",
      "254\n",
      "True\n",
      "186\n",
      "166\n",
      "True\n",
      "187\n",
      "84\n",
      "True\n",
      "188\n",
      "86\n",
      "True\n",
      "189\n",
      "178\n",
      "True\n",
      "190\n",
      "158\n",
      "True\n",
      "191\n",
      "95\n",
      "True\n",
      "192\n",
      "161\n",
      "True\n",
      "193\n",
      "245\n",
      "True\n",
      "194\n",
      "50\n",
      "True\n",
      "195\n",
      "262\n",
      "True\n",
      "196\n",
      "109\n",
      "True\n",
      "197\n",
      "59\n",
      "True\n",
      "198\n",
      "120\n",
      "True\n",
      "199\n",
      "117\n",
      "True\n",
      "200\n",
      "60\n",
      "True\n",
      "201\n",
      "25\n",
      "False\n",
      "202\n",
      "124\n",
      "True\n",
      "203\n",
      "115\n",
      "True\n",
      "204\n",
      "41\n",
      "False\n",
      "205\n",
      "31\n",
      "False\n",
      "206\n",
      "162\n",
      "True\n",
      "207\n",
      "141\n",
      "True\n",
      "208\n",
      "166\n",
      "True\n",
      "209\n",
      "131\n",
      "True\n",
      "210\n",
      "182\n",
      "True\n",
      "211\n",
      "155\n",
      "True\n",
      "212\n",
      "187\n",
      "True\n",
      "213\n",
      "159\n",
      "True\n",
      "214\n",
      "75\n",
      "True\n",
      "215\n",
      "71\n",
      "True\n",
      "216\n",
      "100\n",
      "True\n",
      "217\n",
      "49\n",
      "False\n",
      "218\n",
      "33\n",
      "False\n",
      "219\n",
      "161\n",
      "True\n",
      "220\n",
      "54\n",
      "True\n",
      "221\n",
      "55\n",
      "True\n",
      "222\n",
      "180\n",
      "True\n",
      "223\n",
      "29\n",
      "False\n",
      "224\n",
      "43\n",
      "False\n",
      "225\n",
      "166\n",
      "True\n",
      "226\n",
      "165\n",
      "True\n",
      "227\n",
      "180\n",
      "True\n",
      "228\n",
      "44\n",
      "False\n",
      "229\n",
      "109\n",
      "True\n",
      "230\n",
      "125\n",
      "True\n",
      "231\n",
      "187\n",
      "True\n",
      "232\n",
      "159\n",
      "True\n",
      "233\n",
      "40\n",
      "False\n",
      "234\n",
      "147\n",
      "True\n",
      "235\n",
      "102\n",
      "True\n",
      "236\n",
      "274\n",
      "True\n",
      "237\n",
      "235\n",
      "True\n",
      "238\n",
      "94\n",
      "True\n",
      "239\n",
      "106\n",
      "True\n",
      "240\n",
      "50\n",
      "True\n",
      "241\n",
      "44\n",
      "False\n",
      "242\n",
      "194\n",
      "True\n",
      "243\n",
      "40\n",
      "False\n",
      "244\n",
      "49\n",
      "False\n",
      "245\n",
      "265\n",
      "True\n",
      "246\n",
      "80\n",
      "True\n",
      "247\n",
      "87\n",
      "True\n",
      "248\n",
      "162\n",
      "True\n",
      "249\n",
      "36\n",
      "False\n",
      "250\n",
      "234\n",
      "True\n",
      "251\n",
      "64\n",
      "True\n",
      "252\n",
      "249\n",
      "True\n",
      "253\n",
      "139\n",
      "True\n",
      "254\n",
      "131\n",
      "True\n",
      "255\n",
      "150\n",
      "True\n",
      "256\n",
      "59\n",
      "True\n",
      "257\n",
      "199\n",
      "True\n",
      "258\n",
      "98\n",
      "True\n",
      "259\n",
      "123\n",
      "True\n"
     ]
    }
   ],
   "source": [
    "for id in test['UnitNo'].unique():\n",
    "    print(id)\n",
    "    print(len(test[test['UnitNo']==id]))\n",
    "    print(len(test[test['UnitNo']==id]) >= sequence_length)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(235, 50, 25)"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "seq_array_test_last = [test[test['UnitNo']==id][sequence_cols].values[-sequence_length:] \n",
    "                       for id in test['UnitNo'].unique() if len(test[test['UnitNo']==id]) >= sequence_length]\n",
    "\n",
    "seq_array_test_last = np.asarray(seq_array_test_last).astype(np.float32)\n",
    "seq_array_test_last.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_mask = [len(test[test['UnitNo']==id]) >= sequence_length for id in test['UnitNo'].unique()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "UnitNo\n",
       "1      1\n",
       "2      0\n",
       "3      0\n",
       "4      0\n",
       "5      1\n",
       "6      0\n",
       "7      1\n",
       "8      0\n",
       "9      1\n",
       "10     0\n",
       "11     1\n",
       "12     0\n",
       "13     1\n",
       "14     1\n",
       "15     0\n",
       "16     0\n",
       "17     0\n",
       "18     0\n",
       "19     1\n",
       "20     0\n",
       "21     0\n",
       "22     1\n",
       "23     0\n",
       "24     0\n",
       "25     0\n",
       "26     0\n",
       "27     0\n",
       "28     0\n",
       "29     0\n",
       "30     0\n",
       "      ..\n",
       "230    0\n",
       "231    1\n",
       "232    0\n",
       "233    0\n",
       "234    0\n",
       "235    0\n",
       "236    1\n",
       "237    0\n",
       "238    0\n",
       "239    0\n",
       "240    0\n",
       "241    0\n",
       "242    1\n",
       "243    0\n",
       "244    0\n",
       "245    1\n",
       "246    0\n",
       "247    0\n",
       "248    0\n",
       "249    0\n",
       "250    1\n",
       "251    0\n",
       "252    1\n",
       "253    0\n",
       "254    0\n",
       "255    0\n",
       "256    0\n",
       "257    0\n",
       "258    0\n",
       "259    0\n",
       "Name: label1, Length: 259, dtype: int64"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test.groupby('UnitNo')['label1'].nth(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "UnitNo\n",
       "1      1\n",
       "2      0\n",
       "3      0\n",
       "4      0\n",
       "5      1\n",
       "6      0\n",
       "7      1\n",
       "8      0\n",
       "9      1\n",
       "10     0\n",
       "11     1\n",
       "12     0\n",
       "13     1\n",
       "14     1\n",
       "15     0\n",
       "16     0\n",
       "17     0\n",
       "19     1\n",
       "20     0\n",
       "21     0\n",
       "22     1\n",
       "24     0\n",
       "25     0\n",
       "26     0\n",
       "27     0\n",
       "29     0\n",
       "30     0\n",
       "31     0\n",
       "32     1\n",
       "33     0\n",
       "      ..\n",
       "222    0\n",
       "225    1\n",
       "226    0\n",
       "227    1\n",
       "229    0\n",
       "230    0\n",
       "231    1\n",
       "232    0\n",
       "234    0\n",
       "235    0\n",
       "236    1\n",
       "237    0\n",
       "238    0\n",
       "239    0\n",
       "240    0\n",
       "242    1\n",
       "245    1\n",
       "246    0\n",
       "247    0\n",
       "248    0\n",
       "250    1\n",
       "251    0\n",
       "252    1\n",
       "253    0\n",
       "254    0\n",
       "255    0\n",
       "256    0\n",
       "257    0\n",
       "258    0\n",
       "259    0\n",
       "Name: label1, Length: 235, dtype: int64"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test.groupby('UnitNo')['label1'].nth(-1)[y_mask]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "label_array_test_last = test.groupby('UnitNo')['label1'].nth(-1)[y_mask].values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "235"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "label_array_test_last.shape[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(235, 50, 25)\n",
      "(235, 1)\n"
     ]
    }
   ],
   "source": [
    "label_array_test_last = label_array_test_last.reshape(label_array_test_last.shape[0],1).astype(np.float32)\n",
    "\n",
    "print(seq_array_test_last.shape)\n",
    "print(label_array_test_last.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Accurracy: 0.9702127659574468\n"
     ]
    }
   ],
   "source": [
    "# test metrics\n",
    "scores_test = model.evaluate(seq_array_test_last, label_array_test_last, verbose=2)\n",
    "print('Test Accurracy: {}'.format(scores_test[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion matrix\n",
      "- x-axis is predicted labels.\n",
      "- y-axis is true labels\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[172,   2],\n",
       "       [  5,  56]])"
      ]
     },
     "execution_count": 69,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# make predictions and compute confusion matrix\n",
    "y_pred_test = model.predict_classes(seq_array_test_last)\n",
    "y_true_test = label_array_test_last\n",
    "print('Confusion matrix\\n- x-axis is predicted labels.\\n- y-axis is true labels')\n",
    "cm = confusion_matrix(y_true_test, y_pred_test)\n",
    "cm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Precision:  0.9655172413793104 \n",
      " Test Recall:  0.9180327868852459 \n",
      " Test F1 Score: 0.9411764705882353\n"
     ]
    }
   ],
   "source": [
    "# compute precision and recall\n",
    "precision_test = precision_score(y_true_test, y_pred_test)\n",
    "recall_test = recall_score(y_true_test, y_pred_test)\n",
    "f1_test = 2 * (precision_test * recall_test) / (precision_test + recall_test)\n",
    "print( 'Test Precision: ', precision_test, '\\n', 'Test Recall: ', recall_test, '\\n', 'Test F1 Score:', f1_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.5"
  },
  "toc": {
   "nav_menu": {},
   "number_sections": false,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}