Pandas.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Pandas"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-20T16:35:34.766901Z",
     "start_time": "2019-02-20T16:35:34.521399Z"
    }
   },
   "outputs": [],
   "source": [
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "[Pandas](http://pandas.pydata.org) is a library that provides a set of tools for data analysis (Python Data Analysis Library). "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pandas dataframes can be created by importing a CSV file (or TSV, or JSON, or SQL, etc.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-20T16:35:34.771137Z",
     "start_time": "2019-02-20T16:35:34.768606Z"
    }
   },
   "outputs": [],
   "source": [
    "# df = pd.read_csv(\"file.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pandas dataframes can also be created directly from a dictionary of arrays."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-20T16:35:34.783033Z",
     "start_time": "2019-02-20T16:35:34.774533Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'mean_fit_time': array([0.01142702, 0.01115093, 0.01050925, 0.02054195, 0.07160912,\n",
      "       0.06751108, 0.06907425, 0.06755571, 0.07172804, 0.07606635]), 'std_fit_time': array([0.00053882, 0.00038095, 0.00030056, 0.0004086 , 0.00281235,\n",
      "       0.00261869, 0.00117984, 0.00326013, 0.00573768, 0.0030746 ]), 'mean_score_time': array([0.00030556, 0.00029874, 0.00025778, 0.00029097, 0.00029759,\n",
      "       0.00030074, 0.00031796, 0.00030222, 0.00032935, 0.00035524]), 'std_score_time': array([4.90965395e-05, 1.10162491e-04, 3.79735462e-05, 3.75066905e-05,\n",
      "       4.90331443e-06, 1.30740366e-05, 3.27037297e-05, 1.36716789e-05,\n",
      "       3.10957060e-05, 4.44746199e-05]), 'param_C': masked_array(data=[1e-06, 1e-05, 0.0001, 0.001, 0.01, 0.1, 1.0, 10.0,\n",
      "                   100.0, 1000.0],\n",
      "             mask=[False, False, False, False, False, False, False, False,\n",
      "                   False, False],\n",
      "       fill_value='?',\n",
      "            dtype=object), 'params': [{'C': 1e-06}, {'C': 1e-05}, {'C': 0.0001}, {'C': 0.001}, {'C': 0.01}, {'C': 0.1}, {'C': 1.0}, {'C': 10.0}, {'C': 100.0}, {'C': 1000.0}], 'split0_test_score': array([0.84246575, 0.90753425, 0.93493151, 0.95890411, 0.95547945,\n",
      "       0.94520548, 0.94178082, 0.93835616, 0.92808219, 0.93835616]), 'split1_test_score': array([0.87197232, 0.91349481, 0.94117647, 0.96539792, 0.96193772,\n",
      "       0.97231834, 0.96193772, 0.96539792, 0.94463668, 0.9550173 ]), 'split2_test_score': array([0.90909091, 0.94405594, 0.96503497, 0.96503497, 0.96853147,\n",
      "       0.94755245, 0.93706294, 0.93356643, 0.93356643, 0.93006993]), 'split3_test_score': array([0.84265734, 0.9020979 , 0.94055944, 0.96503497, 0.96153846,\n",
      "       0.96153846, 0.95454545, 0.95104895, 0.94755245, 0.95454545]), 'split4_test_score': array([0.89788732, 0.93309859, 0.96830986, 0.97887324, 0.97183099,\n",
      "       0.95070423, 0.92605634, 0.95422535, 0.94014085, 0.94366197]), 'mean_test_score': array([0.87265136, 0.91997216, 0.94989562, 0.96659708, 0.9638135 ,\n",
      "       0.95546277, 0.94432846, 0.94850383, 0.93876131, 0.94432846]), 'std_test_score': array([0.02746696, 0.01590452, 0.01381196, 0.00656282, 0.0057433 ,\n",
      "       0.0101387 , 0.0126865 , 0.01142328, 0.00715855, 0.00956827]), 'rank_test_score': array([10,  9,  4,  1,  2,  3,  6,  5,  8,  6], dtype=int32), 'split0_train_score': array([0.88296943, 0.92925764, 0.96331878, 0.98253275, 0.99388646,\n",
      "       0.99825328, 0.99737991, 0.99475983, 0.99563319, 0.99650655]), 'split1_train_score': array([0.88066202, 0.93118467, 0.96428571, 0.9825784 , 0.99303136,\n",
      "       0.99651568, 0.99738676, 0.99216028, 0.98954704, 0.98432056]), 'split2_train_score': array([0.86967854, 0.92267593, 0.96003475, 0.98349262, 0.99478714,\n",
      "       0.99826238, 0.99739357, 0.99478714, 0.99565595, 0.99565595]), 'split3_train_score': array([0.88184188, 0.93223284, 0.96090356, 0.98001738, 0.99218071,\n",
      "       0.99652476, 0.99565595, 0.9913119 , 0.98783666, 0.99391833]), 'split4_train_score': array([0.8725065 , 0.92541197, 0.95836947, 0.98178664, 0.99392888,\n",
      "       0.99566349, 0.98091934, 0.99045967, 0.99132697, 0.98785776]), 'mean_train_score': array([0.87753168, 0.92815261, 0.96138246, 0.98208156, 0.99356291,\n",
      "       0.99704392, 0.99374711, 0.99269576, 0.99199996, 0.99165183]), 'std_train_score': array([0.00538275, 0.00359355, 0.00215879, 0.00116527, 0.00088661,\n",
      "       0.00103936, 0.00644882, 0.00177967, 0.00317394, 0.00475144])}\n"
     ]
    }
   ],
   "source": [
    "print(grid_search.cv_results_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-20T16:35:34.824275Z",
     "start_time": "2019-02-20T16:35:34.784975Z"
    }
   },
   "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>mean_fit_time</th>\n",
       "      <th>std_fit_time</th>\n",
       "      <th>mean_score_time</th>\n",
       "      <th>std_score_time</th>\n",
       "      <th>param_C</th>\n",
       "      <th>params</th>\n",
       "      <th>split0_test_score</th>\n",
       "      <th>split1_test_score</th>\n",
       "      <th>split2_test_score</th>\n",
       "      <th>split3_test_score</th>\n",
       "      <th>...</th>\n",
       "      <th>mean_test_score</th>\n",
       "      <th>std_test_score</th>\n",
       "      <th>rank_test_score</th>\n",
       "      <th>split0_train_score</th>\n",
       "      <th>split1_train_score</th>\n",
       "      <th>split2_train_score</th>\n",
       "      <th>split3_train_score</th>\n",
       "      <th>split4_train_score</th>\n",
       "      <th>mean_train_score</th>\n",
       "      <th>std_train_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.011427</td>\n",
       "      <td>0.000539</td>\n",
       "      <td>0.000306</td>\n",
       "      <td>0.000049</td>\n",
       "      <td>1e-06</td>\n",
       "      <td>{'C': 1e-06}</td>\n",
       "      <td>0.842466</td>\n",
       "      <td>0.871972</td>\n",
       "      <td>0.909091</td>\n",
       "      <td>0.842657</td>\n",
       "      <td>...</td>\n",
       "      <td>0.872651</td>\n",
       "      <td>0.027467</td>\n",
       "      <td>10</td>\n",
       "      <td>0.882969</td>\n",
       "      <td>0.880662</td>\n",
       "      <td>0.869679</td>\n",
       "      <td>0.881842</td>\n",
       "      <td>0.872507</td>\n",
       "      <td>0.877532</td>\n",
       "      <td>0.005383</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.011151</td>\n",
       "      <td>0.000381</td>\n",
       "      <td>0.000299</td>\n",
       "      <td>0.000110</td>\n",
       "      <td>1e-05</td>\n",
       "      <td>{'C': 1e-05}</td>\n",
       "      <td>0.907534</td>\n",
       "      <td>0.913495</td>\n",
       "      <td>0.944056</td>\n",
       "      <td>0.902098</td>\n",
       "      <td>...</td>\n",
       "      <td>0.919972</td>\n",
       "      <td>0.015905</td>\n",
       "      <td>9</td>\n",
       "      <td>0.929258</td>\n",
       "      <td>0.931185</td>\n",
       "      <td>0.922676</td>\n",
       "      <td>0.932233</td>\n",
       "      <td>0.925412</td>\n",
       "      <td>0.928153</td>\n",
       "      <td>0.003594</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.010509</td>\n",
       "      <td>0.000301</td>\n",
       "      <td>0.000258</td>\n",
       "      <td>0.000038</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>{'C': 0.0001}</td>\n",
       "      <td>0.934932</td>\n",
       "      <td>0.941176</td>\n",
       "      <td>0.965035</td>\n",
       "      <td>0.940559</td>\n",
       "      <td>...</td>\n",
       "      <td>0.949896</td>\n",
       "      <td>0.013812</td>\n",
       "      <td>4</td>\n",
       "      <td>0.963319</td>\n",
       "      <td>0.964286</td>\n",
       "      <td>0.960035</td>\n",
       "      <td>0.960904</td>\n",
       "      <td>0.958369</td>\n",
       "      <td>0.961382</td>\n",
       "      <td>0.002159</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.020542</td>\n",
       "      <td>0.000409</td>\n",
       "      <td>0.000291</td>\n",
       "      <td>0.000038</td>\n",
       "      <td>0.001</td>\n",
       "      <td>{'C': 0.001}</td>\n",
       "      <td>0.958904</td>\n",
       "      <td>0.965398</td>\n",
       "      <td>0.965035</td>\n",
       "      <td>0.965035</td>\n",
       "      <td>...</td>\n",
       "      <td>0.966597</td>\n",
       "      <td>0.006563</td>\n",
       "      <td>1</td>\n",
       "      <td>0.982533</td>\n",
       "      <td>0.982578</td>\n",
       "      <td>0.983493</td>\n",
       "      <td>0.980017</td>\n",
       "      <td>0.981787</td>\n",
       "      <td>0.982082</td>\n",
       "      <td>0.001165</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.071609</td>\n",
       "      <td>0.002812</td>\n",
       "      <td>0.000298</td>\n",
       "      <td>0.000005</td>\n",
       "      <td>0.01</td>\n",
       "      <td>{'C': 0.01}</td>\n",
       "      <td>0.955479</td>\n",
       "      <td>0.961938</td>\n",
       "      <td>0.968531</td>\n",
       "      <td>0.961538</td>\n",
       "      <td>...</td>\n",
       "      <td>0.963814</td>\n",
       "      <td>0.005743</td>\n",
       "      <td>2</td>\n",
       "      <td>0.993886</td>\n",
       "      <td>0.993031</td>\n",
       "      <td>0.994787</td>\n",
       "      <td>0.992181</td>\n",
       "      <td>0.993929</td>\n",
       "      <td>0.993563</td>\n",
       "      <td>0.000887</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.067511</td>\n",
       "      <td>0.002619</td>\n",
       "      <td>0.000301</td>\n",
       "      <td>0.000013</td>\n",
       "      <td>0.1</td>\n",
       "      <td>{'C': 0.1}</td>\n",
       "      <td>0.945205</td>\n",
       "      <td>0.972318</td>\n",
       "      <td>0.947552</td>\n",
       "      <td>0.961538</td>\n",
       "      <td>...</td>\n",
       "      <td>0.955463</td>\n",
       "      <td>0.010139</td>\n",
       "      <td>3</td>\n",
       "      <td>0.998253</td>\n",
       "      <td>0.996516</td>\n",
       "      <td>0.998262</td>\n",
       "      <td>0.996525</td>\n",
       "      <td>0.995663</td>\n",
       "      <td>0.997044</td>\n",
       "      <td>0.001039</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.069074</td>\n",
       "      <td>0.001180</td>\n",
       "      <td>0.000318</td>\n",
       "      <td>0.000033</td>\n",
       "      <td>1</td>\n",
       "      <td>{'C': 1.0}</td>\n",
       "      <td>0.941781</td>\n",
       "      <td>0.961938</td>\n",
       "      <td>0.937063</td>\n",
       "      <td>0.954545</td>\n",
       "      <td>...</td>\n",
       "      <td>0.944328</td>\n",
       "      <td>0.012686</td>\n",
       "      <td>6</td>\n",
       "      <td>0.997380</td>\n",
       "      <td>0.997387</td>\n",
       "      <td>0.997394</td>\n",
       "      <td>0.995656</td>\n",
       "      <td>0.980919</td>\n",
       "      <td>0.993747</td>\n",
       "      <td>0.006449</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.067556</td>\n",
       "      <td>0.003260</td>\n",
       "      <td>0.000302</td>\n",
       "      <td>0.000014</td>\n",
       "      <td>10</td>\n",
       "      <td>{'C': 10.0}</td>\n",
       "      <td>0.938356</td>\n",
       "      <td>0.965398</td>\n",
       "      <td>0.933566</td>\n",
       "      <td>0.951049</td>\n",
       "      <td>...</td>\n",
       "      <td>0.948504</td>\n",
       "      <td>0.011423</td>\n",
       "      <td>5</td>\n",
       "      <td>0.994760</td>\n",
       "      <td>0.992160</td>\n",
       "      <td>0.994787</td>\n",
       "      <td>0.991312</td>\n",
       "      <td>0.990460</td>\n",
       "      <td>0.992696</td>\n",
       "      <td>0.001780</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.071728</td>\n",
       "      <td>0.005738</td>\n",
       "      <td>0.000329</td>\n",
       "      <td>0.000031</td>\n",
       "      <td>100</td>\n",
       "      <td>{'C': 100.0}</td>\n",
       "      <td>0.928082</td>\n",
       "      <td>0.944637</td>\n",
       "      <td>0.933566</td>\n",
       "      <td>0.947552</td>\n",
       "      <td>...</td>\n",
       "      <td>0.938761</td>\n",
       "      <td>0.007159</td>\n",
       "      <td>8</td>\n",
       "      <td>0.995633</td>\n",
       "      <td>0.989547</td>\n",
       "      <td>0.995656</td>\n",
       "      <td>0.987837</td>\n",
       "      <td>0.991327</td>\n",
       "      <td>0.992000</td>\n",
       "      <td>0.003174</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.076066</td>\n",
       "      <td>0.003075</td>\n",
       "      <td>0.000355</td>\n",
       "      <td>0.000044</td>\n",
       "      <td>1000</td>\n",
       "      <td>{'C': 1000.0}</td>\n",
       "      <td>0.938356</td>\n",
       "      <td>0.955017</td>\n",
       "      <td>0.930070</td>\n",
       "      <td>0.954545</td>\n",
       "      <td>...</td>\n",
       "      <td>0.944328</td>\n",
       "      <td>0.009568</td>\n",
       "      <td>6</td>\n",
       "      <td>0.996507</td>\n",
       "      <td>0.984321</td>\n",
       "      <td>0.995656</td>\n",
       "      <td>0.993918</td>\n",
       "      <td>0.987858</td>\n",
       "      <td>0.991652</td>\n",
       "      <td>0.004751</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>10 rows × 21 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   mean_fit_time  std_fit_time  mean_score_time  std_score_time param_C  \\\n",
       "0       0.011427      0.000539         0.000306        0.000049   1e-06   \n",
       "1       0.011151      0.000381         0.000299        0.000110   1e-05   \n",
       "2       0.010509      0.000301         0.000258        0.000038  0.0001   \n",
       "3       0.020542      0.000409         0.000291        0.000038   0.001   \n",
       "4       0.071609      0.002812         0.000298        0.000005    0.01   \n",
       "5       0.067511      0.002619         0.000301        0.000013     0.1   \n",
       "6       0.069074      0.001180         0.000318        0.000033       1   \n",
       "7       0.067556      0.003260         0.000302        0.000014      10   \n",
       "8       0.071728      0.005738         0.000329        0.000031     100   \n",
       "9       0.076066      0.003075         0.000355        0.000044    1000   \n",
       "\n",
       "          params  split0_test_score  split1_test_score  split2_test_score  \\\n",
       "0   {'C': 1e-06}           0.842466           0.871972           0.909091   \n",
       "1   {'C': 1e-05}           0.907534           0.913495           0.944056   \n",
       "2  {'C': 0.0001}           0.934932           0.941176           0.965035   \n",
       "3   {'C': 0.001}           0.958904           0.965398           0.965035   \n",
       "4    {'C': 0.01}           0.955479           0.961938           0.968531   \n",
       "5     {'C': 0.1}           0.945205           0.972318           0.947552   \n",
       "6     {'C': 1.0}           0.941781           0.961938           0.937063   \n",
       "7    {'C': 10.0}           0.938356           0.965398           0.933566   \n",
       "8   {'C': 100.0}           0.928082           0.944637           0.933566   \n",
       "9  {'C': 1000.0}           0.938356           0.955017           0.930070   \n",
       "\n",
       "   split3_test_score       ...         mean_test_score  std_test_score  \\\n",
       "0           0.842657       ...                0.872651        0.027467   \n",
       "1           0.902098       ...                0.919972        0.015905   \n",
       "2           0.940559       ...                0.949896        0.013812   \n",
       "3           0.965035       ...                0.966597        0.006563   \n",
       "4           0.961538       ...                0.963814        0.005743   \n",
       "5           0.961538       ...                0.955463        0.010139   \n",
       "6           0.954545       ...                0.944328        0.012686   \n",
       "7           0.951049       ...                0.948504        0.011423   \n",
       "8           0.947552       ...                0.938761        0.007159   \n",
       "9           0.954545       ...                0.944328        0.009568   \n",
       "\n",
       "   rank_test_score  split0_train_score  split1_train_score  \\\n",
       "0               10            0.882969            0.880662   \n",
       "1                9            0.929258            0.931185   \n",
       "2                4            0.963319            0.964286   \n",
       "3                1            0.982533            0.982578   \n",
       "4                2            0.993886            0.993031   \n",
       "5                3            0.998253            0.996516   \n",
       "6                6            0.997380            0.997387   \n",
       "7                5            0.994760            0.992160   \n",
       "8                8            0.995633            0.989547   \n",
       "9                6            0.996507            0.984321   \n",
       "\n",
       "   split2_train_score  split3_train_score  split4_train_score  \\\n",
       "0            0.869679            0.881842            0.872507   \n",
       "1            0.922676            0.932233            0.925412   \n",
       "2            0.960035            0.960904            0.958369   \n",
       "3            0.983493            0.980017            0.981787   \n",
       "4            0.994787            0.992181            0.993929   \n",
       "5            0.998262            0.996525            0.995663   \n",
       "6            0.997394            0.995656            0.980919   \n",
       "7            0.994787            0.991312            0.990460   \n",
       "8            0.995656            0.987837            0.991327   \n",
       "9            0.995656            0.993918            0.987858   \n",
       "\n",
       "   mean_train_score  std_train_score  \n",
       "0          0.877532         0.005383  \n",
       "1          0.928153         0.003594  \n",
       "2          0.961382         0.002159  \n",
       "3          0.982082         0.001165  \n",
       "4          0.993563         0.000887  \n",
       "5          0.997044         0.001039  \n",
       "6          0.993747         0.006449  \n",
       "7          0.992696         0.001780  \n",
       "8          0.992000         0.003174  \n",
       "9          0.991652         0.004751  \n",
       "\n",
       "[10 rows x 21 columns]"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.DataFrame(grid_search.cv_results_)\n",
    "df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pandas columns are also Numpy arrays, so they obey to the same indexing magic"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-20T16:35:34.851638Z",
     "start_time": "2019-02-20T16:35:34.826575Z"
    }
   },
   "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>mean_fit_time</th>\n",
       "      <th>std_fit_time</th>\n",
       "      <th>mean_score_time</th>\n",
       "      <th>std_score_time</th>\n",
       "      <th>param_C</th>\n",
       "      <th>params</th>\n",
       "      <th>split0_test_score</th>\n",
       "      <th>split1_test_score</th>\n",
       "      <th>split2_test_score</th>\n",
       "      <th>split3_test_score</th>\n",
       "      <th>...</th>\n",
       "      <th>mean_test_score</th>\n",
       "      <th>std_test_score</th>\n",
       "      <th>rank_test_score</th>\n",
       "      <th>split0_train_score</th>\n",
       "      <th>split1_train_score</th>\n",
       "      <th>split2_train_score</th>\n",
       "      <th>split3_train_score</th>\n",
       "      <th>split4_train_score</th>\n",
       "      <th>mean_train_score</th>\n",
       "      <th>std_train_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.011427</td>\n",
       "      <td>0.000539</td>\n",
       "      <td>0.000306</td>\n",
       "      <td>0.000049</td>\n",
       "      <td>1e-06</td>\n",
       "      <td>{'C': 1e-06}</td>\n",
       "      <td>0.842466</td>\n",
       "      <td>0.871972</td>\n",
       "      <td>0.909091</td>\n",
       "      <td>0.842657</td>\n",
       "      <td>...</td>\n",
       "      <td>0.872651</td>\n",
       "      <td>0.027467</td>\n",
       "      <td>10</td>\n",
       "      <td>0.882969</td>\n",
       "      <td>0.880662</td>\n",
       "      <td>0.869679</td>\n",
       "      <td>0.881842</td>\n",
       "      <td>0.872507</td>\n",
       "      <td>0.877532</td>\n",
       "      <td>0.005383</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.011151</td>\n",
       "      <td>0.000381</td>\n",
       "      <td>0.000299</td>\n",
       "      <td>0.000110</td>\n",
       "      <td>1e-05</td>\n",
       "      <td>{'C': 1e-05}</td>\n",
       "      <td>0.907534</td>\n",
       "      <td>0.913495</td>\n",
       "      <td>0.944056</td>\n",
       "      <td>0.902098</td>\n",
       "      <td>...</td>\n",
       "      <td>0.919972</td>\n",
       "      <td>0.015905</td>\n",
       "      <td>9</td>\n",
       "      <td>0.929258</td>\n",
       "      <td>0.931185</td>\n",
       "      <td>0.922676</td>\n",
       "      <td>0.932233</td>\n",
       "      <td>0.925412</td>\n",
       "      <td>0.928153</td>\n",
       "      <td>0.003594</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.010509</td>\n",
       "      <td>0.000301</td>\n",
       "      <td>0.000258</td>\n",
       "      <td>0.000038</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>{'C': 0.0001}</td>\n",
       "      <td>0.934932</td>\n",
       "      <td>0.941176</td>\n",
       "      <td>0.965035</td>\n",
       "      <td>0.940559</td>\n",
       "      <td>...</td>\n",
       "      <td>0.949896</td>\n",
       "      <td>0.013812</td>\n",
       "      <td>4</td>\n",
       "      <td>0.963319</td>\n",
       "      <td>0.964286</td>\n",
       "      <td>0.960035</td>\n",
       "      <td>0.960904</td>\n",
       "      <td>0.958369</td>\n",
       "      <td>0.961382</td>\n",
       "      <td>0.002159</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.020542</td>\n",
       "      <td>0.000409</td>\n",
       "      <td>0.000291</td>\n",
       "      <td>0.000038</td>\n",
       "      <td>0.001</td>\n",
       "      <td>{'C': 0.001}</td>\n",
       "      <td>0.958904</td>\n",
       "      <td>0.965398</td>\n",
       "      <td>0.965035</td>\n",
       "      <td>0.965035</td>\n",
       "      <td>...</td>\n",
       "      <td>0.966597</td>\n",
       "      <td>0.006563</td>\n",
       "      <td>1</td>\n",
       "      <td>0.982533</td>\n",
       "      <td>0.982578</td>\n",
       "      <td>0.983493</td>\n",
       "      <td>0.980017</td>\n",
       "      <td>0.981787</td>\n",
       "      <td>0.982082</td>\n",
       "      <td>0.001165</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>4 rows × 21 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   mean_fit_time  std_fit_time  mean_score_time  std_score_time param_C  \\\n",
       "0       0.011427      0.000539         0.000306        0.000049   1e-06   \n",
       "1       0.011151      0.000381         0.000299        0.000110   1e-05   \n",
       "2       0.010509      0.000301         0.000258        0.000038  0.0001   \n",
       "3       0.020542      0.000409         0.000291        0.000038   0.001   \n",
       "\n",
       "          params  split0_test_score  split1_test_score  split2_test_score  \\\n",
       "0   {'C': 1e-06}           0.842466           0.871972           0.909091   \n",
       "1   {'C': 1e-05}           0.907534           0.913495           0.944056   \n",
       "2  {'C': 0.0001}           0.934932           0.941176           0.965035   \n",
       "3   {'C': 0.001}           0.958904           0.965398           0.965035   \n",
       "\n",
       "   split3_test_score       ...         mean_test_score  std_test_score  \\\n",
       "0           0.842657       ...                0.872651        0.027467   \n",
       "1           0.902098       ...                0.919972        0.015905   \n",
       "2           0.940559       ...                0.949896        0.013812   \n",
       "3           0.965035       ...                0.966597        0.006563   \n",
       "\n",
       "   rank_test_score  split0_train_score  split1_train_score  \\\n",
       "0               10            0.882969            0.880662   \n",
       "1                9            0.929258            0.931185   \n",
       "2                4            0.963319            0.964286   \n",
       "3                1            0.982533            0.982578   \n",
       "\n",
       "   split2_train_score  split3_train_score  split4_train_score  \\\n",
       "0            0.869679            0.881842            0.872507   \n",
       "1            0.922676            0.932233            0.925412   \n",
       "2            0.960035            0.960904            0.958369   \n",
       "3            0.983493            0.980017            0.981787   \n",
       "\n",
       "   mean_train_score  std_train_score  \n",
       "0          0.877532         0.005383  \n",
       "1          0.928153         0.003594  \n",
       "2          0.961382         0.002159  \n",
       "3          0.982082         0.001165  \n",
       "\n",
       "[4 rows x 21 columns]"
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[df['param_C'] < 0.01]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "They also provide most functionality you would expect as database user (`df.sort_values`, `df.groupby`, `df.join`, `df.concat`, etc.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-20T16:35:35.071086Z",
     "start_time": "2019-02-20T16:35:34.853988Z"
    },
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots()\n",
    "ax.plot(df['mean_test_score'], label=\"validation error\")\n",
    "ax.plot(df['mean_train_score'], label=\"training error\")\n",
    "ax.set_xticklabels(df['param_C'])\n",
    "ax.set_xlabel(\"C\")\n",
    "ax.set_ylabel(\"Accuracy\")\n",
    "ax.legend(loc='best');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-19T20:36:16.683621Z",
     "start_time": "2019-02-19T20:36:16.679694Z"
    }
   },
   "source": [
    "## Other packages "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "Other packages that didn't make the cut:\n",
    "- [Plotly](https://plot.ly) and [Seaborn](http://seaborn.pydata.org): two other plotting libraries\n",
    "- [Scipy](https://www.scipy.org): a science library built on top of Numpy\n",
    "- [Scrapy](https://www.scipy.org): a web crawling library\n",
    "- [pdb](https://docs.python.org/3/library/pdb.html): a debugger for python (not ML-specific but terribly useful)\n",
    "- [tqdm](https://github.com/tqdm/tqdm): a progress bar (not ML-specific)\n",
    "\n",
    "Next time:\n",
    "- [Pytorch](https://pytorch.org)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Pandas for Stock Market Analysis"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create Dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "x = np.random.uniform(low=-5.0, high=5.0, size=(5,))  # uniform distribution given by p(x) = 1/(b-a), [a, b)\n",
    "y = - 5 * x - x ** 2 + 1 * x ** 3 - 0.1 * x ** 4 + np.random.normal(loc=0, scale=5,  size=(5,))\n",
    "\n",
    "df = pd.DataFrame(np.column_stack((x, y)), columns=['x', 'y'])\n",
    "df.to_csv('regression_train_assignment2017.csv')  # to save dataframe as csv\n",
    "\n",
    "# ### Get the training data as numpy arrays\n",
    "data_train = pd.read_csv('regression_test_assignment2017.csv')\n",
    "x_train = data_train['x'].to_numpy()\n",
    "y_train = data_train['y'].to_numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'pd' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-1-8460206bbe6b>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      7\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      8\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0m__name__\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m\"__main__\"\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 9\u001b[0;31m     \u001b[0mtest_run\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     10\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-1-8460206bbe6b>\u001b[0m in \u001b[0;36mtest_run\u001b[0;34m()\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mtest_run\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      2\u001b[0m     \u001b[0;34m\"\"\"Function called by Test Run.\"\"\"\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 3\u001b[0;31m     \u001b[0mdf\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mpd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mread_csv\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"data/AAPL.csv\"\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;31m# wont work since there is no AAPL.csv\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      4\u001b[0m     \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtail\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m  \u001b[0;31m# Prints last 5 rows of the data frame\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m     \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhead\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m  \u001b[0;31m# Prints first 5 rows of the data frame\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'pd' is not defined"
     ]
    }
   ],
   "source": [
    "def test_run():\n",
    "    \"\"\"Function called by Test Run.\"\"\"\n",
    "    df = pd.read_csv(\"data/AAPL.csv\") # wont work since there is no AAPL.csv\n",
    "    print(df.tail())  # Prints last 5 rows of the data frame\n",
    "    print(df.head())  # Prints first 5 rows of the data frame\n",
    "    print(df[10:21])  # Slicing for data from 10 - 20\n",
    "    df['High'].plot() # Needs Matplotlib.pyplot as plt\n",
    "    df[['Low','High']].plot() # To plot two datas in the same graph\n",
    "    plt.show() \n",
    "\n",
    "if __name__ == \"__main__\":\n",
    "    test_run()\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_mean_volume(symbol):\n",
    "    \"\"\"Return the mean volume for stock indicated by symbol.\n",
    "    \n",
    "    Note: Data for a stock is stored in file: data/<symbol>.csv\n",
    "    \"\"\"\n",
    "    df = pd.read_csv(\"data/{}.csv\".format(symbol))  # read in data\n",
    "    # TODO: Compute and return the mean volume for this stock\n",
    "    # return df['Volume'].max()\n",
    "    return df['Volume'].mean()\n",
    "\n",
    "\n",
    "def test_run():\n",
    "    \"\"\"Function called by Test Run.\"\"\"\n",
    "    for symbol in ['AAPL', 'IBM']:\n",
    "        print \"Mean Volume\"\n",
    "        print symbol, get_mean_volume(symbol)\n",
    "\n",
    "\n",
    "if __name__ == \"__main__\":\n",
    "    test_run()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\"\"\"Utility functions\"\"\"\n",
    "\n",
    "import os\n",
    "import pandas as pd\n",
    "\n",
    "def symbol_to_path(symbol, base_dir=\"data\"):\n",
    "    \"\"\"Return CSV file path given ticker symbol, for example IBM. The return will be \n",
    "    'path_to_dir/data/IBM.csv' \"\"\"\n",
    "    return os.path.join(base_dir, \"{}.csv\".format(str(symbol)))\n",
    "\n",
    "\n",
    "def get_data(symbols, dates):\n",
    "    \"\"\"Read stock data (adjusted close) for given symbols from CSV files.\"\"\"\n",
    "    df = pd.DataFrame(index=dates)\n",
    "    if 'SPY' not in symbols:  # add SPY for reference, if absent\n",
    "        symbols.insert(0, 'SPY')\n",
    "    print(df.head())\n",
    "    for symbol in symbols:\n",
    "        # TODO: Read and join data for each symbol\n",
    "        path = symbol_to_path(symbol)\n",
    "        df_temp = pd.read_csv(path, index_col='Date', \n",
    "                              parse_dates=True, usecols=['Date', 'Adj Close'],\n",
    "                             na_values=['nan'])\n",
    "        df_temp = df_temp.rename(columns={'Adj Close':symbol})\n",
    "        df = df.join(df_temp)\n",
    "        if symbol == 'SPY': # drop dates when SPY did not trade\n",
    "            df = df.dropna(subset=[\"SPY\"])\n",
    "\n",
    "    return df\n",
    "\n",
    "def plot_data(df, title=\"Stock prices\"):\n",
    "    \"\"\"Plot stock prices with a custom title and meaningful axis labels.\"\"\"\n",
    "    ax = df.plot(title=title, fontsize=12)\n",
    "    ax.set_xlabel(\"Date\")\n",
    "    ax.set_ylabel(\"Price\")\n",
    "    plt.show()\n",
    "    \n",
    "def plot_selected(df, columns, start_index, end_index):\n",
    "    \"\"\"Plot the desired columns over index values in the given range.\"\"\"\n",
    "    # TODO: Your code here\n",
    "    data = df.ix[start_index:end_index, columns]\n",
    "    plot_data(data)\n",
    "    \n",
    "def normalize(df):\n",
    "    return df/df.ix[0, :]\n",
    "\n",
    "def statistics(df):\n",
    "    mean = df.mean()\n",
    "    std_dev = df.std()\n",
    "    median = df.median()\n",
    "    \"TO get all values in a dataframe\"\n",
    "    values = df.values()\n",
    "\n",
    "\n",
    "def test_run():\n",
    "    # Define a date range\n",
    "    dates = pd.date_range('2010-01-01', '2010-12-31') # jan to dec\n",
    "\n",
    "    # Choose stock symbols to read\n",
    "    symbols = ['GOOG', 'IBM', 'GLD']\n",
    "    # df = pd.read_csv('data/IBM.csv')\n",
    "    # print(df.head())\n",
    "    # Get stock data\n",
    "    df = get_data(symbols, dates)\n",
    "    \n",
    "    \"\"\" Slicing data\"\"\"\n",
    "    df.ix['2010-01-01':'2010-01-31'] # full jan\n",
    "    print(df.ix['2010-01-01':'2010-01-31', ['GOOG', 'GLD']]) # full jan for goog and gld\n",
    "    print df\n",
    "\n",
    "\n",
    "if __name__ == \"__main__\":\n",
    "    test_run()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Rolling Mean, Rolling std_dev and Bollinger bands"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Bollinger bands are found out as ```roll_mean +/- 2 * roll_std_dev```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\"\"\"Bollinger Bands.\"\"\"\n",
    "\n",
    "import os\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "def get_rolling_mean(values, window):\n",
    "    \"\"\"Return rolling mean of given values, using specified window size.\"\"\"\n",
    "    # return pd.rolling_mean(values, window=window)\n",
    "    return values.rolling(center=False, window=window).mean()\n",
    "\n",
    "\n",
    "def get_rolling_std(values, window):\n",
    "    \"\"\"Return rolling standard deviation of given values, using specified window size.\"\"\"\n",
    "    # TODO: Compute and return rolling standard deviation\n",
    "    # return pd.rolling_std(values, window=window)\n",
    "    return values.rolling(center=False, window=window).std()\n",
    "\n",
    "def get_bollinger_bands(rm, rstd):\n",
    "    \"\"\"Return upper and lower Bollinger Bands.\"\"\"\n",
    "    # TODO: Compute upper_band and lower_band\n",
    "    upper_band = rm + 2*rstd\n",
    "    lower_band = rm - 2*rstd\n",
    "    return upper_band, lower_band\n",
    "\n",
    "def compute_daily_returns(df):\n",
    "    \"\"\"Using Numpy\"\"\"\n",
    "    Ndaily_returns = df.copy()\n",
    "    Ndaily_returns = (df[1:]/df[:-1].values) - 1\n",
    "    Ndaily_returns.ix[0, :] = 0  # set daily returns of row 0 to 0\n",
    "    \n",
    "    \"\"\"Using pandas\"\"\"\n",
    "    Pdaily_returns = (df/df.shift(1)) - 1\n",
    "    Pdaily_returns.ix[0, :] = 0  # set daily returns of row 0 to 0\n",
    "    return daily_returns\n",
    "\n",
    "def cumulative_returns(df, time_point):\n",
    "    cum_returns = (df[time_point, :]/df[0, :]) - 1\n",
    "    cum_returns.ix[0, :] = 0  # set daily returns of row 0 to 0\n",
    "    return cum_returns\n",
    "\n",
    "def test_run():\n",
    "    # Read data\n",
    "    dates = pd.date_range('2012-01-01', '2012-12-31')\n",
    "    symbols = ['SPY']\n",
    "    df = get_data(symbols, dates)\n",
    "\n",
    "    # Compute Bollinger Bands\n",
    "    # 1. Compute rolling mean\n",
    "    rm_SPY = get_rolling_mean(df['SPY'], window=20)\n",
    "\n",
    "    # 2. Compute rolling standard deviation\n",
    "    rstd_SPY = get_rolling_std(df['SPY'], window=20)\n",
    "\n",
    "    # 3. Compute upper and lower bands\n",
    "    upper_band, lower_band = get_bollinger_bands(rm_SPY, rstd_SPY)\n",
    "    \n",
    "    # Plot raw SPY values, rolling mean and Bollinger Bands\n",
    "    ax = df['SPY'].plot(title=\"Bollinger Bands\", label='SPY')\n",
    "    rm_SPY.plot(label='Rolling mean', ax=ax)\n",
    "    upper_band.plot(label='upper band', ax=ax)\n",
    "    lower_band.plot(label='lower band', ax=ax)\n",
    "\n",
    "    # Add axis labels and legend\n",
    "    ax.set_xlabel(\"Date\")\n",
    "    ax.set_ylabel(\"Price\")\n",
    "    ax.legend(loc='upper left')\n",
    "    plt.show()\n",
    "\n",
    "\n",
    "if __name__ == \"__main__\":\n",
    "    test_run()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Fill Missing Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\"\"\"Fill missing values\"\"\"\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import os\n",
    "\n",
    "def fill_missing_values(df_data):\n",
    "    \"\"\"Fill missing values in data frame, in place.\"\"\"\n",
    "    ##########################################################\n",
    "    # TODO: Your code here (DO NOT modify anything else)\n",
    "    df_data.fillna(method='ffill', inplace=True)\n",
    "    df_data.fillna(method='bfill', inplace=True)\n",
    "    ##########################################################\n",
    "\n",
    "\n",
    "def symbol_to_path(symbol, base_dir=\"data\"):\n",
    "    \"\"\"Return CSV file path given ticker symbol.\"\"\"\n",
    "    return os.path.join(base_dir, \"{}.csv\".format(str(symbol)))\n",
    "\n",
    "def test_run():\n",
    "    \"\"\"Function called by Test Run.\"\"\"\n",
    "    # Read data\n",
    "    symbol_list = [\"JAVA\", \"FAKE1\", \"FAKE2\"]  # list of symbols\n",
    "    start_date = \"2005-12-31\"\n",
    "    end_date = \"2014-12-07\"\n",
    "    dates = pd.date_range(start_date, end_date)  # date range as index\n",
    "    df_data = get_data(symbol_list, dates)  # get data for each symbol\n",
    "\n",
    "    # Fill missing values\n",
    "    fill_missing_values(df_data)\n",
    "\n",
    "    # Plot\n",
    "    plot_data(df_data)\n",
    "\n",
    "\n",
    "if __name__ == \"__main__\":\n",
    "    test_run()\n"
   ]
  }
 ],
 "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": "3.6.8"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {
    "height": "calc(100% - 180px)",
    "left": "10px",
    "top": "150px",
    "width": "164px"
   },
   "toc_section_display": true,
   "toc_window_display": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}