3.5 - RF - PNU Repeated Random Subsampling Random Search.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Setup, matplotlib inline, automatically reload libraries on every evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import scipy as sp\n",
    "pd.options.display.max_rows = 400\n",
    "pd.options.display.max_columns = 400\n",
    "%matplotlib inline\n",
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from epiml.loadepiml import LoadEpiml, load_search, save_search\n",
    "from epiml.epimlsklearn.epimlmetrics import pu_scorer, prior_squared_error_scorer_015, brier_score_labeled_loss_scorer, \\\n",
    "    f1_assumed_scorer, f1_labeled_scorer, report_metrics, f1_assumed_beta10_scorer, pu_mix_assumed_f1beta10_scorer\n",
    "from epiml.semisuperhelper import SemiSupervisedHelper\n",
    "from epiml.epimlsklearn.pnuwrapper import PNUWrapper\n",
    "from epiml.epimlsklearn.jsearchcv import JRandomSearchCV, extract_score_grid\n",
    "from epiml.epimlsklearn.nestedcross import NestedCV, rerun_nested_for_scoring\n",
    "from epiml.epimlsklearn.frankenscorer import FrankenScorer, extract_scores_from_nested\n",
    "from epiml.epimlsklearn.repeatedsampling import RepeatedRandomSubSampler"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "path = \"C:\\\\Data\\\\010317\\\\membership14_final_0103.txt\"\n",
    "lc = LoadEpiml(path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.base import clone"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(lc.X, lc.y, test_size=0.2, random_state=771, stratify=lc.y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# REPEATED SUB SAMPLER"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Set up PNU Wrapper with Random Forest, then JSearchCV, then NestedCV"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "rf = RandomForestClassifier()\n",
    "rep = RepeatedRandomSubSampler(base_estimator=rf, voting='thresh', verbose=1, random_state=83)\n",
    "pnu = PNUWrapper(base_estimator=rep, num_unlabeled=1.0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Set up randomized search parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "rf_param_search = {'base_estimator__base_estimator__bootstrap': [True, False],\n",
    " 'base_estimator__base_estimator__class_weight': [None,'balanced','balanced_subsample'],\n",
    " 'base_estimator__base_estimator__criterion': ['gini','entropy'],\n",
    " 'base_estimator__base_estimator__max_depth': [None] + list(range(2,100)),\n",
    " 'base_estimator__base_estimator__max_features': ['sqrt','log2',None] + list(range(5,100)),\n",
    " 'base_estimator__base_estimator__min_samples_leaf': [1,2,3,4,5,6,7,8,9,10,15,20,25,30,35,40,45,50,75,100],\n",
    " 'base_estimator__base_estimator__min_samples_split':[2,0.005,0.01,0.015,0.02,0.025,0.03,0.035,0.04,\n",
    "                                                        0.045,0.05,0.07,0.09,0.1,0.12,0.15,0.17,0.2,0.25],\n",
    " 'base_estimator__base_estimator__n_estimators': sp.stats.randint(low=10, high=300),\n",
    " 'base_estimator__sample_imbalance': sp.stats.uniform(loc=0.1, scale=0.9),\n",
    " 'pu_learning': [True, False]\n",
    "                  }"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### notice random_state is set in jsearch, this is so that the same random parameters are searched for each outer fold, sort of like grid search"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "jsearch = JRandomSearchCV(pnu, rf_param_search, n_iter=60, scoring=FrankenScorer(decision_score='assumed_f1beta10'),\n",
    "                                n_jobs=-1, cv=3, verbose=1, pre_dispatch=8, random_state=77)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3x3 (x60) nested cross validation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "nested_cross = NestedCV(jsearch, scoring=FrankenScorer(decision_score='pu_mix_assumed_f1beta10'), cv=3, random_state=77)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Score the nested cross"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[CV]  ................................................................\n",
      "Fitting 3 folds for each of 60 candidates, totalling 180 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done  42 tasks      | elapsed: 222.2min\n",
      "[Parallel(n_jobs=-1)]: Done 180 out of 180 | elapsed: 1005.4min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "generating 74 samples of indices to use to train multiple estimators, sized 4299 elements with last being 1884 elements\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done  74 out of  74 | elapsed:  4.4min finished\n",
      "[Parallel(n_jobs=1)]: Done  74 out of  74 | elapsed:  3.8min finished\n",
      "[Parallel(n_jobs=1)]: Done  74 out of  74 | elapsed:  3.9min finished\n",
      "[Parallel(n_jobs=1)]: Done  74 out of  74 | elapsed:  7.9min finished\n",
      "[Parallel(n_jobs=1)]: Done  74 out of  74 | elapsed:  7.9min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[CV]  , score=82.5854652613 score_data={'labeled_acc': 0.8224852071005917, 'labeled_prec': 0.9094567404426559, 'labeled_recall': 0.81149012567324952, 'labeled_f1': 0.85768500948766602, 'labeled_roc_auc': 0.82762006283662481, 'labeled_avg_prec': 0.92260361057274554, 'labeled_brier': 0.14938732268854843, 'labeled_brier_pos': 0.16735656238839849, 'labeled_brier_neg': 0.11463431396349126, 'confusion_matrix_lab': array([[243,  45],\n",
      "       [105, 452]]), 'pr_one_unlabeled': 0.059910006428112277, 'assumed_brier': 0.043873714556292492, 'assumed_brier_neg': 0.043285665803759782, 'assumed_f1': 0.11238189955246147, 'assumed_f1beta10': 0.72249038568059887, 'confusion_matrix_un': array([[109928,   7035],\n",
      "       [   105,    452]]), 'pu_score': 10.336426693220341, 'pu_mix_assumed_f1beta10': 82.585465261280234, 'SCORE': 82.585465261280234}, total=1017.6min\n",
      "[Parallel(n_jobs=1)]: Done   1 out of   1 | elapsed: 1033.4min remaining:    0.0s\n",
      "[CV]  ................................................................\n",
      "Fitting 3 folds for each of 60 candidates, totalling 180 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done  42 tasks      | elapsed: 224.3min\n",
      "[Parallel(n_jobs=-1)]: Done 180 out of 180 | elapsed: 1012.2min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "generating 74 samples of indices to use to train multiple estimators, sized 4299 elements with last being 1884 elements\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done  74 out of  74 | elapsed:  4.6min finished\n",
      "[Parallel(n_jobs=1)]: Done  74 out of  74 | elapsed:  3.9min finished\n",
      "[Parallel(n_jobs=1)]: Done  74 out of  74 | elapsed:  3.9min finished\n",
      "[Parallel(n_jobs=1)]: Done  74 out of  74 | elapsed:  8.0min finished\n",
      "[Parallel(n_jobs=1)]: Done  74 out of  74 | elapsed:  8.0min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[CV]  , score=81.8979342765 score_data={'labeled_acc': 0.81990521327014221, 'labeled_prec': 0.92099792099792099, 'labeled_recall': 0.79533213644524237, 'labeled_f1': 0.8535645472061657, 'labeled_roc_auc': 0.83146397763028668, 'labeled_avg_prec': 0.92570057374527825, 'labeled_brier': 0.15900027409777365, 'labeled_brier_pos': 0.18425020768865785, 'labeled_brier_neg': 0.10999604758166738, 'confusion_matrix_lab': array([[249,  38],\n",
      "       [114, 443]]), 'pr_one_unlabeled': 0.056138847225198199, 'assumed_brier': 0.042591307186482255, 'assumed_brier_neg': 0.041916694854445255, 'assumed_f1': 0.11676331049024777, 'assumed_f1beta10': 0.71325182126859121, 'confusion_matrix_un': array([[110374,   6588],\n",
      "       [   114,    443]]), 'pu_score': 10.572752149664066, 'pu_mix_assumed_f1beta10': 81.897934276523188, 'SCORE': 81.897934276523188}, total=1024.7min\n",
      "[Parallel(n_jobs=1)]: Done   2 out of   2 | elapsed: 2074.2min remaining:    0.0s\n",
      "[CV]  ................................................................\n",
      "Fitting 3 folds for each of 60 candidates, totalling 180 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done  42 tasks      | elapsed: 223.2min\n",
      "[Parallel(n_jobs=-1)]: Done 180 out of 180 | elapsed: 1012.2min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "generating 156 samples of indices to use to train multiple estimators, sized 2621 elements with last being 1454 elements\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done 156 out of 156 | elapsed:  4.3min finished\n",
      "[Parallel(n_jobs=1)]: Done 156 out of 156 | elapsed:  5.0min finished\n",
      "[Parallel(n_jobs=1)]: Done 156 out of 156 | elapsed:  5.0min finished\n",
      "[Parallel(n_jobs=1)]: Done 156 out of 156 | elapsed: 10.2min finished\n",
      "[Parallel(n_jobs=1)]: Done 156 out of 156 | elapsed: 10.2min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[CV]  , score=80.3540053814 score_data={'labeled_acc': 0.83392645314353497, 'labeled_prec': 0.91434262948207168, 'labeled_recall': 0.82553956834532372, 'labeled_f1': 0.86767485822306234, 'labeled_roc_auc': 0.8378568921865992, 'labeled_avg_prec': 0.92747372050325882, 'labeled_brier': 0.12806831101477065, 'labeled_brier_pos': 0.11153994584788486, 'labeled_brier_neg': 0.16008841914295363, 'confusion_matrix_lab': array([[244,  43],\n",
      "       [ 97, 459]]), 'pr_one_unlabeled': 0.074103913468296284, 'assumed_brier': 0.072174460224645631, 'assumed_brier_neg': 0.071987327676133542, 'assumed_f1': 0.094600164880461662, 'assumed_f1beta10': 0.7159912275282635, 'confusion_matrix_un': array([[108272,   8689],\n",
      "       [    97,    459]]), 'pu_score': 8.7548826285566168, 'pu_mix_assumed_f1beta10': 80.354005381382962, 'SCORE': 80.354005381382962}, total=1026.5min\n",
      "[Parallel(n_jobs=1)]: Done   3 out of   3 | elapsed: 3121.1min remaining:    0.0s\n",
      "[Parallel(n_jobs=1)]: Done   3 out of   3 | elapsed: 3121.1min finished\n"
     ]
    }
   ],
   "source": [
    "scores = nested_cross.score(X_train.values, y=y_train.values, verbose=100, pre_dispatch=8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "save_search(nested_cross, './res/nested_cross_repreated_rf_large.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "nested_cross = load_search('./res/nested_cross_repreated_rf_large.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "assumed_brier                                                      0.0528798\n",
       "assumed_brier_neg                                                  0.0523966\n",
       "assumed_f1                                                          0.107915\n",
       "assumed_f1beta10                                                    0.717244\n",
       "confusion_matrix_lab       [[245.333333333, 42.0], [105.333333333, 451.33...\n",
       "confusion_matrix_un        [[109524.666667, 7437.33333333], [105.33333333...\n",
       "fn_confusion_matrix_lab                                              105.333\n",
       "fn_confusion_matrix_un                                               105.333\n",
       "fp_confusion_matrix_lab                                                   42\n",
       "fp_confusion_matrix_un                                               7437.33\n",
       "labeled_acc                                                         0.825439\n",
       "labeled_avg_prec                                                    0.925259\n",
       "labeled_brier                                                       0.145485\n",
       "labeled_brier_neg                                                    0.12824\n",
       "labeled_brier_pos                                                   0.154382\n",
       "labeled_f1                                                          0.859641\n",
       "labeled_prec                                                        0.914932\n",
       "labeled_recall                                                      0.810787\n",
       "labeled_roc_auc                                                     0.832314\n",
       "pr_one_unlabeled                                                   0.0633843\n",
       "pu_mix_assumed_f1beta10                                              81.6125\n",
       "pu_score                                                             9.88802\n",
       "tn_confusion_matrix_lab                                              245.333\n",
       "tn_confusion_matrix_un                                                109525\n",
       "tp_confusion_matrix_lab                                              451.333\n",
       "tp_confusion_matrix_un                                               451.333\n",
       "dtype: object"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "extract_scores_from_nested(nested_cross.test_score_datas_).mean().sort_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[51, 51, 7]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nested_cross.best_idxs_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[{'base_estimator__base_estimator__bootstrap': False,\n",
       "  'base_estimator__base_estimator__class_weight': None,\n",
       "  'base_estimator__base_estimator__criterion': 'gini',\n",
       "  'base_estimator__base_estimator__max_depth': 47,\n",
       "  'base_estimator__base_estimator__max_features': 49,\n",
       "  'base_estimator__base_estimator__min_samples_leaf': 6,\n",
       "  'base_estimator__base_estimator__min_samples_split': 0.02,\n",
       "  'base_estimator__base_estimator__n_estimators': 177,\n",
       "  'base_estimator__sample_imbalance': 0.34927465901033783,\n",
       "  'pu_learning': True},\n",
       " {'base_estimator__base_estimator__bootstrap': False,\n",
       "  'base_estimator__base_estimator__class_weight': None,\n",
       "  'base_estimator__base_estimator__criterion': 'gini',\n",
       "  'base_estimator__base_estimator__max_depth': 47,\n",
       "  'base_estimator__base_estimator__max_features': 49,\n",
       "  'base_estimator__base_estimator__min_samples_leaf': 6,\n",
       "  'base_estimator__base_estimator__min_samples_split': 0.02,\n",
       "  'base_estimator__base_estimator__n_estimators': 177,\n",
       "  'base_estimator__sample_imbalance': 0.34927465901033783,\n",
       "  'pu_learning': True},\n",
       " {'base_estimator__base_estimator__bootstrap': False,\n",
       "  'base_estimator__base_estimator__class_weight': 'balanced_subsample',\n",
       "  'base_estimator__base_estimator__criterion': 'entropy',\n",
       "  'base_estimator__base_estimator__max_depth': 64,\n",
       "  'base_estimator__base_estimator__max_features': 55,\n",
       "  'base_estimator__base_estimator__min_samples_leaf': 2,\n",
       "  'base_estimator__base_estimator__min_samples_split': 0.005,\n",
       "  'base_estimator__base_estimator__n_estimators': 108,\n",
       "  'base_estimator__sample_imbalance': 0.73873322019678955,\n",
       "  'pu_learning': False}]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nested_cross.best_params_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Now that we have comparable, nested search scores, lets do a search on the whole 80% training set and use the parameters found"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "jsearch_train = JRandomSearchCV(pnu, rf_param_search, n_iter=60, scoring=FrankenScorer(decision_score='pu_mix_assumed_f1beta10'),\n",
    "                                n_jobs=-1, cv=3, verbose=1, pre_dispatch=8, random_state=77)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 60 candidates, totalling 180 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done  42 tasks      | elapsed: 347.2min\n",
      "[Parallel(n_jobs=-1)]: Done 180 out of 180 | elapsed: 1584.4min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "generating 74 samples of indices to use to train multiple estimators, sized 6451 elements with last being 2681 elements\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done  74 out of  74 | elapsed:  7.8min finished\n"
     ]
    }
   ],
   "source": [
    "jsearch_fit = jsearch_train.fit(X_train.values, y_train.values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "save_search(jsearch_fit, './res/jsearch_only.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "jsearch_score_grid = extract_score_grid(jsearch_fit)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "best_idx = jsearch_score_grid.mean_pu_mix_assumed_f1beta10_test.idxmax()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model6_params = pd.DataFrame(jsearch_fit.cv_results_).params[best_idx]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'base_estimator__base_estimator__bootstrap': False,\n",
       " 'base_estimator__base_estimator__class_weight': None,\n",
       " 'base_estimator__base_estimator__criterion': 'gini',\n",
       " 'base_estimator__base_estimator__max_depth': 47,\n",
       " 'base_estimator__base_estimator__max_features': 49,\n",
       " 'base_estimator__base_estimator__min_samples_leaf': 6,\n",
       " 'base_estimator__base_estimator__min_samples_split': 0.02,\n",
       " 'base_estimator__base_estimator__n_estimators': 177,\n",
       " 'base_estimator__sample_imbalance': 0.34927465901033783,\n",
       " 'pu_learning': True}"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model6_params"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "## This is model 3 in the manuscript, lets retrain on whole training set and then test on 20% test set!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "PNUWrapper(base_estimator=RepeatedRandomSubSampler(base_estimator=RandomForestClassifier(bootstrap=False, class_weight=None, criterion='gini',\n",
       "            max_depth=47, max_features=49, max_leaf_nodes=None,\n",
       "            min_impurity_split=1e-07, min_samples_leaf=6,\n",
       "            min_samples_split=0.02, min_weight...     random_state=83, sample_imbalance=0.34927465901033783,\n",
       "             verbose=1, voting='thresh'),\n",
       "      num_unlabeled=1.0, pu_learning=True, random_state=None,\n",
       "      threshold_set_pct=None)"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pnu_test = clone(pnu)\n",
    "pnu_test.set_params(**model6_params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "generating 74 samples of indices to use to train multiple estimators, sized 6451 elements with last being 2681 elements\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done  74 out of  74 | elapsed:  8.0min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "PNUWrapper(base_estimator=RepeatedRandomSubSampler(base_estimator=RandomForestClassifier(bootstrap=False, class_weight=None, criterion='gini',\n",
       "            max_depth=47, max_features=49, max_leaf_nodes=None,\n",
       "            min_impurity_split=1e-07, min_samples_leaf=6,\n",
       "            min_samples_split=0.02, min_weight...     random_state=83, sample_imbalance=0.34927465901033783,\n",
       "             verbose=1, voting='thresh'),\n",
       "      num_unlabeled=1.0, pu_learning=True, random_state=None,\n",
       "      threshold_set_pct=None)"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pnu_test.fit(X_train.values, y_train.values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done  74 out of  74 | elapsed:  3.4min finished\n",
      "[Parallel(n_jobs=1)]: Done  74 out of  74 | elapsed:  3.3min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "({'SCORE': 0.85496183206106879,\n",
       "  'assumed_brier': 0.042653797253650108,\n",
       "  'assumed_brier_neg': 0.042055028774489925,\n",
       "  'assumed_f1': 0.1169305724725944,\n",
       "  'assumed_f1beta10': 0.72006620127734511,\n",
       "  'confusion_matrix_lab': array([[183,  32],\n",
       "         [ 82, 336]]),\n",
       "  'confusion_matrix_un': array([[82729,  4993],\n",
       "         [   82,   336]]),\n",
       "  'labeled_acc': 0.81990521327014221,\n",
       "  'labeled_avg_prec': 0.92320654679803116,\n",
       "  'labeled_brier': 0.14829033063607508,\n",
       "  'labeled_brier_neg': 0.10936429540334816,\n",
       "  'labeled_brier_pos': 0.16831209516965467,\n",
       "  'labeled_f1': 0.85496183206106879,\n",
       "  'labeled_prec': 0.91304347826086951,\n",
       "  'labeled_recall': 0.80382775119617222,\n",
       "  'labeled_roc_auc': 0.82749527094692332,\n",
       "  'pr_one_unlabeled': 0.056692607448546976,\n",
       "  'pu_mix_assumed_f1beta10': 82.693558799848489,\n",
       "  'pu_score': 10.686938672113984},\n",
       " 0.85496183206106879)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "FrankenScorer()(pnu_test, X_test.values, y_test.values)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Feature importance for 80/20"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Importance</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG4_Sum</th>\n",
       "      <td>39.101</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG5_Sum</th>\n",
       "      <td>23.853</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat58_Sum</th>\n",
       "      <td>5.287</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG9_Sum</th>\n",
       "      <td>3.897</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat87_Sum</th>\n",
       "      <td>3.179</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG75_Sum</th>\n",
       "      <td>2.428</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat54_Sum</th>\n",
       "      <td>1.146</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>age</th>\n",
       "      <td>1.093</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG43_Sum</th>\n",
       "      <td>0.910</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG42_Sum</th>\n",
       "      <td>0.886</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG2_Sum</th>\n",
       "      <td>0.640</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG47_Sum</th>\n",
       "      <td>0.630</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG6_Sum</th>\n",
       "      <td>0.621</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG38_Sum</th>\n",
       "      <td>0.592</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG48_Sum</th>\n",
       "      <td>0.531</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG9_Sum</th>\n",
       "      <td>0.489</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG1_Sum</th>\n",
       "      <td>0.443</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat18_Sum</th>\n",
       "      <td>0.425</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat85_Sum</th>\n",
       "      <td>0.419</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG3_Sum</th>\n",
       "      <td>0.371</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG16_Sum</th>\n",
       "      <td>0.367</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG7_Sum</th>\n",
       "      <td>0.357</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG18_Sum</th>\n",
       "      <td>0.328</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG53_Sum</th>\n",
       "      <td>0.328</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat15_Sum</th>\n",
       "      <td>0.317</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG15_Sum</th>\n",
       "      <td>0.312</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG80_Sum</th>\n",
       "      <td>0.309</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG6_Sum</th>\n",
       "      <td>0.302</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG22_Sum</th>\n",
       "      <td>0.300</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG11_Sum</th>\n",
       "      <td>0.299</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG77_Sum</th>\n",
       "      <td>0.298</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG48_Sum</th>\n",
       "      <td>0.286</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG12_Sum</th>\n",
       "      <td>0.285</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat16_Sum</th>\n",
       "      <td>0.282</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat45_Sum</th>\n",
       "      <td>0.272</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG11_Sum</th>\n",
       "      <td>0.258</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat11_Sum</th>\n",
       "      <td>0.225</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG41_Sum</th>\n",
       "      <td>0.224</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG54_Sum</th>\n",
       "      <td>0.206</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG1_Sum</th>\n",
       "      <td>0.204</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG67_Sum</th>\n",
       "      <td>0.203</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG60_Sum</th>\n",
       "      <td>0.203</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG19_Sum</th>\n",
       "      <td>0.200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG8_Sum</th>\n",
       "      <td>0.183</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG31_Sum</th>\n",
       "      <td>0.177</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG76_Sum</th>\n",
       "      <td>0.177</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat40_Sum</th>\n",
       "      <td>0.177</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat20_Sum</th>\n",
       "      <td>0.160</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG49_Sum</th>\n",
       "      <td>0.151</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat98_Sum</th>\n",
       "      <td>0.150</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG85_Sum</th>\n",
       "      <td>0.141</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG4_Sum</th>\n",
       "      <td>0.124</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG10_Sum</th>\n",
       "      <td>0.120</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat70_Sum</th>\n",
       "      <td>0.120</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG72_Sum</th>\n",
       "      <td>0.118</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat82_Sum</th>\n",
       "      <td>0.117</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG49_Sum</th>\n",
       "      <td>0.110</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat29_Sum</th>\n",
       "      <td>0.106</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG71_Sum</th>\n",
       "      <td>0.102</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Gender_M</th>\n",
       "      <td>0.100</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat63_Sum</th>\n",
       "      <td>0.099</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat13_Sum</th>\n",
       "      <td>0.096</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat76_Sum</th>\n",
       "      <td>0.095</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat22_Sum</th>\n",
       "      <td>0.093</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG55_Sum</th>\n",
       "      <td>0.091</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat59_Sum</th>\n",
       "      <td>0.089</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat77_Sum</th>\n",
       "      <td>0.088</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat21_Sum</th>\n",
       "      <td>0.085</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG70_Sum</th>\n",
       "      <td>0.084</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat49_Sum</th>\n",
       "      <td>0.084</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG38_Sum</th>\n",
       "      <td>0.080</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG21_Sum</th>\n",
       "      <td>0.079</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat72_Sum</th>\n",
       "      <td>0.075</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG20_Sum</th>\n",
       "      <td>0.075</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG45_Sum</th>\n",
       "      <td>0.074</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG46_Sum</th>\n",
       "      <td>0.072</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat23_Sum</th>\n",
       "      <td>0.067</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG12_Sum</th>\n",
       "      <td>0.067</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG34_Sum</th>\n",
       "      <td>0.065</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat44_Sum</th>\n",
       "      <td>0.064</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG24_Sum</th>\n",
       "      <td>0.059</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG35_Sum</th>\n",
       "      <td>0.057</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG12_Sum</th>\n",
       "      <td>0.057</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG10_Sum</th>\n",
       "      <td>0.055</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG51_Sum</th>\n",
       "      <td>0.053</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG67_Sum</th>\n",
       "      <td>0.052</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG17_Sum</th>\n",
       "      <td>0.050</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG86_Sum</th>\n",
       "      <td>0.048</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG15_Sum</th>\n",
       "      <td>0.048</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG66_Sum</th>\n",
       "      <td>0.046</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat31_Sum</th>\n",
       "      <td>0.046</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat66_Sum</th>\n",
       "      <td>0.045</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG73_Sum</th>\n",
       "      <td>0.045</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat83_Sum</th>\n",
       "      <td>0.045</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat79_Sum</th>\n",
       "      <td>0.044</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat6_Sum</th>\n",
       "      <td>0.044</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat60_Sum</th>\n",
       "      <td>0.043</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG23_Sum</th>\n",
       "      <td>0.043</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat100_Sum</th>\n",
       "      <td>0.041</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG79_Sum</th>\n",
       "      <td>0.040</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG45_Sum</th>\n",
       "      <td>0.039</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG30_Sum</th>\n",
       "      <td>0.039</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG34_Sum</th>\n",
       "      <td>0.038</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat39_Sum</th>\n",
       "      <td>0.038</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat53_Sum</th>\n",
       "      <td>0.037</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG84_Sum</th>\n",
       "      <td>0.037</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG68_Sum</th>\n",
       "      <td>0.037</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG62_Sum</th>\n",
       "      <td>0.037</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat86_Sum</th>\n",
       "      <td>0.037</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat5_Sum</th>\n",
       "      <td>0.037</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat43_Sum</th>\n",
       "      <td>0.035</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG22_Sum</th>\n",
       "      <td>0.035</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG74_Sum</th>\n",
       "      <td>0.034</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat28_Sum</th>\n",
       "      <td>0.033</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG36_Sum</th>\n",
       "      <td>0.032</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG63_Sum</th>\n",
       "      <td>0.032</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG83_Sum</th>\n",
       "      <td>0.032</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG10_Sum</th>\n",
       "      <td>0.030</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG21_Sum</th>\n",
       "      <td>0.030</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG52_Sum</th>\n",
       "      <td>0.030</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG14_Sum</th>\n",
       "      <td>0.029</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG35_Sum</th>\n",
       "      <td>0.029</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat30_Sum</th>\n",
       "      <td>0.028</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG63_Sum</th>\n",
       "      <td>0.028</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat14_Sum</th>\n",
       "      <td>0.028</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat37_Sum</th>\n",
       "      <td>0.028</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat104_Sum</th>\n",
       "      <td>0.028</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG37_Sum</th>\n",
       "      <td>0.026</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat42_Sum</th>\n",
       "      <td>0.025</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG28_Sum</th>\n",
       "      <td>0.024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat105_Sum</th>\n",
       "      <td>0.024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat33_Sum</th>\n",
       "      <td>0.024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat35_Sum</th>\n",
       "      <td>0.023</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat19_Sum</th>\n",
       "      <td>0.023</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG20_Sum</th>\n",
       "      <td>0.023</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG47_Sum</th>\n",
       "      <td>0.022</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat38_Sum</th>\n",
       "      <td>0.021</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG50_Sum</th>\n",
       "      <td>0.020</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG2_Sum</th>\n",
       "      <td>0.020</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG40_Sum</th>\n",
       "      <td>0.020</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat78_Sum</th>\n",
       "      <td>0.019</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat41_Sum</th>\n",
       "      <td>0.019</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG24_Sum</th>\n",
       "      <td>0.019</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat68_Sum</th>\n",
       "      <td>0.018</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG69_Sum</th>\n",
       "      <td>0.018</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat61_Sum</th>\n",
       "      <td>0.017</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG25_Sum</th>\n",
       "      <td>0.017</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG19_Sum</th>\n",
       "      <td>0.016</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat67_Sum</th>\n",
       "      <td>0.016</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG39_Sum</th>\n",
       "      <td>0.015</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG17_Sum</th>\n",
       "      <td>0.015</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat71_Sum</th>\n",
       "      <td>0.013</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat97_Sum</th>\n",
       "      <td>0.013</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG32_Sum</th>\n",
       "      <td>0.013</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat103_Sum</th>\n",
       "      <td>0.013</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG82_Sum</th>\n",
       "      <td>0.012</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG3_Sum</th>\n",
       "      <td>0.012</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat74_Sum</th>\n",
       "      <td>0.012</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG37_Sum</th>\n",
       "      <td>0.012</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat4_Sum</th>\n",
       "      <td>0.011</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG3_Sum</th>\n",
       "      <td>0.011</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG61_Sum</th>\n",
       "      <td>0.010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG26_Sum</th>\n",
       "      <td>0.010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat1_Sum</th>\n",
       "      <td>0.010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat27_Sum</th>\n",
       "      <td>0.010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat96_Sum</th>\n",
       "      <td>0.010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat80_Sum</th>\n",
       "      <td>0.010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG31_Sum</th>\n",
       "      <td>0.009</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG28_Sum</th>\n",
       "      <td>0.009</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG65_Sum</th>\n",
       "      <td>0.009</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG2_Sum</th>\n",
       "      <td>0.008</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat26_Sum</th>\n",
       "      <td>0.008</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG33_Sum</th>\n",
       "      <td>0.008</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat7_Sum</th>\n",
       "      <td>0.008</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG8_Sum</th>\n",
       "      <td>0.007</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG64_Sum</th>\n",
       "      <td>0.007</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat106_Sum</th>\n",
       "      <td>0.007</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG13_Sum</th>\n",
       "      <td>0.006</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG32_Sum</th>\n",
       "      <td>0.006</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG7_Sum</th>\n",
       "      <td>0.006</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat10_Sum</th>\n",
       "      <td>0.006</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG37_Sum</th>\n",
       "      <td>0.006</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat36_Sum</th>\n",
       "      <td>0.006</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG70_Sum</th>\n",
       "      <td>0.006</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG34_Sum</th>\n",
       "      <td>0.006</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG5_Sum</th>\n",
       "      <td>0.005</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG33_Sum</th>\n",
       "      <td>0.005</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG6_Sum</th>\n",
       "      <td>0.005</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat94_Sum</th>\n",
       "      <td>0.005</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG66_Sum</th>\n",
       "      <td>0.005</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG41_Sum</th>\n",
       "      <td>0.004</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG87_Sum</th>\n",
       "      <td>0.004</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat47_Sum</th>\n",
       "      <td>0.003</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat95_Sum</th>\n",
       "      <td>0.003</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG43_Sum</th>\n",
       "      <td>0.003</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG26_Sum</th>\n",
       "      <td>0.003</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PROC_FLAG4_Sum</th>\n",
       "      <td>0.003</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat102_Sum</th>\n",
       "      <td>0.003</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG36_Sum</th>\n",
       "      <td>0.003</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG29_Sum</th>\n",
       "      <td>0.003</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat48_Sum</th>\n",
       "      <td>0.003</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG55_Sum</th>\n",
       "      <td>0.002</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG68_Sum</th>\n",
       "      <td>0.002</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG29_Sum</th>\n",
       "      <td>0.002</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG78_Sum</th>\n",
       "      <td>0.002</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG7_Sum</th>\n",
       "      <td>0.002</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG48_Sum</th>\n",
       "      <td>0.002</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat88_Sum</th>\n",
       "      <td>0.001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat24_Sum</th>\n",
       "      <td>0.001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG46_Sum</th>\n",
       "      <td>0.001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG61_Sum</th>\n",
       "      <td>0.001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG1_Sum</th>\n",
       "      <td>0.001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG81_Sum</th>\n",
       "      <td>0.001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG33_Sum</th>\n",
       "      <td>0.001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG13_Sum</th>\n",
       "      <td>0.001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG60_Sum</th>\n",
       "      <td>0.001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG29_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG16_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG69_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat65_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat34_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG26_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG16_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG23_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG40_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat12_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG39_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG4_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG36_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat73_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG15_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat17_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG18_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG27_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG32_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG35_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG44_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG30_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG49_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG22_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PROC_FLAG5_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat101_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat9_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat2_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG9_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG5_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat3_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat89_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat90_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat91_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat8_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG17_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG14_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat32_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG25_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat99_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG11_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CPT_FLAG8_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG62_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PROC_FLAG2_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DIAG_FLAG27_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat107_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat84_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat56_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat57_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat69_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG52_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG51_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG50_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat62_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat64_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat55_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG47_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG46_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat25_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG42_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG58_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG53_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat52_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat50_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG54_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat75_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG56_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG57_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG21_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ndc_cat46_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>REVCODE_FLAG59_Sum</th>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    Importance\n",
       "DIAG_FLAG4_Sum          39.101\n",
       "DIAG_FLAG5_Sum          23.853\n",
       "ndc_cat58_Sum            5.287\n",
       "CPT_FLAG9_Sum            3.897\n",
       "ndc_cat87_Sum            3.179\n",
       "DIAG_FLAG75_Sum          2.428\n",
       "ndc_cat54_Sum            1.146\n",
       "age                      1.093\n",
       "CPT_FLAG43_Sum           0.910\n",
       "DIAG_FLAG42_Sum          0.886\n",
       "DIAG_FLAG2_Sum           0.640\n",
       "CPT_FLAG47_Sum           0.630\n",
       "DIAG_FLAG6_Sum           0.621\n",
       "DIAG_FLAG38_Sum          0.592\n",
       "DIAG_FLAG48_Sum          0.531\n",
       "DIAG_FLAG9_Sum           0.489\n",
       "CPT_FLAG1_Sum            0.443\n",
       "ndc_cat18_Sum            0.425\n",
       "ndc_cat85_Sum            0.419\n",
       "DIAG_FLAG3_Sum           0.371\n",
       "REVCODE_FLAG16_Sum       0.367\n",
       "DIAG_FLAG7_Sum           0.357\n",
       "CPT_FLAG18_Sum           0.328\n",
       "DIAG_FLAG53_Sum          0.328\n",
       "ndc_cat15_Sum            0.317\n",
       "REVCODE_FLAG15_Sum       0.312\n",
       "DIAG_FLAG80_Sum          0.309\n",
       "CPT_FLAG6_Sum            0.302\n",
       "REVCODE_FLAG22_Sum       0.300\n",
       "DIAG_FLAG11_Sum          0.299\n",
       "DIAG_FLAG77_Sum          0.298\n",
       "CPT_FLAG48_Sum           0.286\n",
       "CPT_FLAG12_Sum           0.285\n",
       "ndc_cat16_Sum            0.282\n",
       "ndc_cat45_Sum            0.272\n",
       "REVCODE_FLAG11_Sum       0.258\n",
       "ndc_cat11_Sum            0.225\n",
       "DIAG_FLAG41_Sum          0.224\n",
       "DIAG_FLAG54_Sum          0.206\n",
       "DIAG_FLAG1_Sum           0.204\n",
       "REVCODE_FLAG67_Sum       0.203\n",
       "DIAG_FLAG60_Sum          0.203\n",
       "REVCODE_FLAG19_Sum       0.200\n",
       "DIAG_FLAG8_Sum           0.183\n",
       "DIAG_FLAG31_Sum          0.177\n",
       "DIAG_FLAG76_Sum          0.177\n",
       "ndc_cat40_Sum            0.177\n",
       "ndc_cat20_Sum            0.160\n",
       "DIAG_FLAG49_Sum          0.151\n",
       "ndc_cat98_Sum            0.150\n",
       "DIAG_FLAG85_Sum          0.141\n",
       "CPT_FLAG4_Sum            0.124\n",
       "DIAG_FLAG10_Sum          0.120\n",
       "ndc_cat70_Sum            0.120\n",
       "DIAG_FLAG72_Sum          0.118\n",
       "ndc_cat82_Sum            0.117\n",
       "CPT_FLAG49_Sum           0.110\n",
       "ndc_cat29_Sum            0.106\n",
       "DIAG_FLAG71_Sum          0.102\n",
       "Gender_M                 0.100\n",
       "ndc_cat63_Sum            0.099\n",
       "ndc_cat13_Sum            0.096\n",
       "ndc_cat76_Sum            0.095\n",
       "ndc_cat22_Sum            0.093\n",
       "DIAG_FLAG55_Sum          0.091\n",
       "ndc_cat59_Sum            0.089\n",
       "ndc_cat77_Sum            0.088\n",
       "ndc_cat21_Sum            0.085\n",
       "DIAG_FLAG70_Sum          0.084\n",
       "ndc_cat49_Sum            0.084\n",
       "REVCODE_FLAG38_Sum       0.080\n",
       "CPT_FLAG21_Sum           0.079\n",
       "ndc_cat72_Sum            0.075\n",
       "CPT_FLAG20_Sum           0.075\n",
       "DIAG_FLAG45_Sum          0.074\n",
       "CPT_FLAG46_Sum           0.072\n",
       "ndc_cat23_Sum            0.067\n",
       "DIAG_FLAG12_Sum          0.067\n",
       "DIAG_FLAG34_Sum          0.065\n",
       "ndc_cat44_Sum            0.064\n",
       "DIAG_FLAG24_Sum          0.059\n",
       "CPT_FLAG35_Sum           0.057\n",
       "REVCODE_FLAG12_Sum       0.057\n",
       "REVCODE_FLAG10_Sum       0.055\n",
       "DIAG_FLAG51_Sum          0.053\n",
       "DIAG_FLAG67_Sum          0.052\n",
       "REVCODE_FLAG17_Sum       0.050\n",
       "DIAG_FLAG86_Sum          0.048\n",
       "DIAG_FLAG15_Sum          0.048\n",
       "DIAG_FLAG66_Sum          0.046\n",
       "ndc_cat31_Sum            0.046\n",
       "ndc_cat66_Sum            0.045\n",
       "DIAG_FLAG73_Sum          0.045\n",
       "ndc_cat83_Sum            0.045\n",
       "ndc_cat79_Sum            0.044\n",
       "ndc_cat6_Sum             0.044\n",
       "ndc_cat60_Sum            0.043\n",
       "REVCODE_FLAG23_Sum       0.043\n",
       "ndc_cat100_Sum           0.041\n",
       "DIAG_FLAG79_Sum          0.040\n",
       "CPT_FLAG45_Sum           0.039\n",
       "DIAG_FLAG30_Sum          0.039\n",
       "REVCODE_FLAG34_Sum       0.038\n",
       "ndc_cat39_Sum            0.038\n",
       "ndc_cat53_Sum            0.037\n",
       "DIAG_FLAG84_Sum          0.037\n",
       "DIAG_FLAG68_Sum          0.037\n",
       "REVCODE_FLAG62_Sum       0.037\n",
       "ndc_cat86_Sum            0.037\n",
       "ndc_cat5_Sum             0.037\n",
       "ndc_cat43_Sum            0.035\n",
       "CPT_FLAG22_Sum           0.035\n",
       "DIAG_FLAG74_Sum          0.034\n",
       "ndc_cat28_Sum            0.033\n",
       "CPT_FLAG36_Sum           0.032\n",
       "DIAG_FLAG63_Sum          0.032\n",
       "DIAG_FLAG83_Sum          0.032\n",
       "CPT_FLAG10_Sum           0.030\n",
       "DIAG_FLAG21_Sum          0.030\n",
       "DIAG_FLAG52_Sum          0.030\n",
       "REVCODE_FLAG14_Sum       0.029\n",
       "DIAG_FLAG35_Sum          0.029\n",
       "ndc_cat30_Sum            0.028\n",
       "REVCODE_FLAG63_Sum       0.028\n",
       "ndc_cat14_Sum            0.028\n",
       "ndc_cat37_Sum            0.028\n",
       "ndc_cat104_Sum           0.028\n",
       "DIAG_FLAG37_Sum          0.026\n",
       "ndc_cat42_Sum            0.025\n",
       "REVCODE_FLAG28_Sum       0.024\n",
       "ndc_cat105_Sum           0.024\n",
       "ndc_cat33_Sum            0.024\n",
       "ndc_cat35_Sum            0.023\n",
       "ndc_cat19_Sum            0.023\n",
       "REVCODE_FLAG20_Sum       0.023\n",
       "DIAG_FLAG47_Sum          0.022\n",
       "ndc_cat38_Sum            0.021\n",
       "DIAG_FLAG50_Sum          0.020\n",
       "REVCODE_FLAG2_Sum        0.020\n",
       "DIAG_FLAG40_Sum          0.020\n",
       "ndc_cat78_Sum            0.019\n",
       "ndc_cat41_Sum            0.019\n",
       "REVCODE_FLAG24_Sum       0.019\n",
       "ndc_cat68_Sum            0.018\n",
       "DIAG_FLAG69_Sum          0.018\n",
       "ndc_cat61_Sum            0.017\n",
       "REVCODE_FLAG25_Sum       0.017\n",
       "CPT_FLAG19_Sum           0.016\n",
       "ndc_cat67_Sum            0.016\n",
       "REVCODE_FLAG39_Sum       0.015\n",
       "DIAG_FLAG17_Sum          0.015\n",
       "ndc_cat71_Sum            0.013\n",
       "ndc_cat97_Sum            0.013\n",
       "CPT_FLAG32_Sum           0.013\n",
       "ndc_cat103_Sum           0.013\n",
       "DIAG_FLAG82_Sum          0.012\n",
       "REVCODE_FLAG3_Sum        0.012\n",
       "ndc_cat74_Sum            0.012\n",
       "CPT_FLAG37_Sum           0.012\n",
       "ndc_cat4_Sum             0.011\n",
       "CPT_FLAG3_Sum            0.011\n",
       "DIAG_FLAG61_Sum          0.010\n",
       "CPT_FLAG26_Sum           0.010\n",
       "ndc_cat1_Sum             0.010\n",
       "ndc_cat27_Sum            0.010\n",
       "ndc_cat96_Sum            0.010\n",
       "ndc_cat80_Sum            0.010\n",
       "REVCODE_FLAG31_Sum       0.009\n",
       "DIAG_FLAG28_Sum          0.009\n",
       "REVCODE_FLAG65_Sum       0.009\n",
       "CPT_FLAG2_Sum            0.008\n",
       "ndc_cat26_Sum            0.008\n",
       "REVCODE_FLAG33_Sum       0.008\n",
       "ndc_cat7_Sum             0.008\n",
       "REVCODE_FLAG8_Sum        0.007\n",
       "REVCODE_FLAG64_Sum       0.007\n",
       "ndc_cat106_Sum           0.007\n",
       "CPT_FLAG13_Sum           0.006\n",
       "DIAG_FLAG32_Sum          0.006\n",
       "CPT_FLAG7_Sum            0.006\n",
       "ndc_cat10_Sum            0.006\n",
       "REVCODE_FLAG37_Sum       0.006\n",
       "ndc_cat36_Sum            0.006\n",
       "REVCODE_FLAG70_Sum       0.006\n",
       "CPT_FLAG34_Sum           0.006\n",
       "CPT_FLAG5_Sum            0.005\n",
       "CPT_FLAG33_Sum           0.005\n",
       "REVCODE_FLAG6_Sum        0.005\n",
       "ndc_cat94_Sum            0.005\n",
       "REVCODE_FLAG66_Sum       0.005\n",
       "REVCODE_FLAG41_Sum       0.004\n",
       "DIAG_FLAG87_Sum          0.004\n",
       "ndc_cat47_Sum            0.003\n",
       "ndc_cat95_Sum            0.003\n",
       "REVCODE_FLAG43_Sum       0.003\n",
       "REVCODE_FLAG26_Sum       0.003\n",
       "PROC_FLAG4_Sum           0.003\n",
       "ndc_cat102_Sum           0.003\n",
       "REVCODE_FLAG36_Sum       0.003\n",
       "DIAG_FLAG29_Sum          0.003\n",
       "ndc_cat48_Sum            0.003\n",
       "REVCODE_FLAG55_Sum       0.002\n",
       "REVCODE_FLAG68_Sum       0.002\n",
       "CPT_FLAG29_Sum           0.002\n",
       "DIAG_FLAG78_Sum          0.002\n",
       "REVCODE_FLAG7_Sum        0.002\n",
       "REVCODE_FLAG48_Sum       0.002\n",
       "ndc_cat88_Sum            0.001\n",
       "ndc_cat24_Sum            0.001\n",
       "DIAG_FLAG46_Sum          0.001\n",
       "REVCODE_FLAG61_Sum       0.001\n",
       "REVCODE_FLAG1_Sum        0.001\n",
       "DIAG_FLAG81_Sum          0.001\n",
       "DIAG_FLAG33_Sum          0.001\n",
       "REVCODE_FLAG13_Sum       0.001\n",
       "REVCODE_FLAG60_Sum       0.001\n",
       "REVCODE_FLAG29_Sum       0.000\n",
       "CPT_FLAG16_Sum           0.000\n",
       "REVCODE_FLAG69_Sum       0.000\n",
       "ndc_cat65_Sum            0.000\n",
       "ndc_cat34_Sum            0.000\n",
       "DIAG_FLAG26_Sum          0.000\n",
       "DIAG_FLAG16_Sum          0.000\n",
       "DIAG_FLAG23_Sum          0.000\n",
       "REVCODE_FLAG40_Sum       0.000\n",
       "ndc_cat12_Sum            0.000\n",
       "DIAG_FLAG39_Sum          0.000\n",
       "REVCODE_FLAG4_Sum        0.000\n",
       "DIAG_FLAG36_Sum          0.000\n",
       "ndc_cat73_Sum            0.000\n",
       "CPT_FLAG15_Sum           0.000\n",
       "ndc_cat17_Sum            0.000\n",
       "REVCODE_FLAG18_Sum       0.000\n",
       "REVCODE_FLAG27_Sum       0.000\n",
       "REVCODE_FLAG32_Sum       0.000\n",
       "REVCODE_FLAG35_Sum       0.000\n",
       "REVCODE_FLAG44_Sum       0.000\n",
       "REVCODE_FLAG30_Sum       0.000\n",
       "REVCODE_FLAG49_Sum       0.000\n",
       "DIAG_FLAG22_Sum          0.000\n",
       "PROC_FLAG5_Sum           0.000\n",
       "ndc_cat101_Sum           0.000\n",
       "ndc_cat9_Sum             0.000\n",
       "ndc_cat2_Sum             0.000\n",
       "REVCODE_FLAG9_Sum        0.000\n",
       "REVCODE_FLAG5_Sum        0.000\n",
       "ndc_cat3_Sum             0.000\n",
       "ndc_cat89_Sum            0.000\n",
       "ndc_cat90_Sum            0.000\n",
       "ndc_cat91_Sum            0.000\n",
       "ndc_cat8_Sum             0.000\n",
       "CPT_FLAG17_Sum           0.000\n",
       "CPT_FLAG14_Sum           0.000\n",
       "ndc_cat32_Sum            0.000\n",
       "DIAG_FLAG25_Sum          0.000\n",
       "ndc_cat99_Sum            0.000\n",
       "CPT_FLAG11_Sum           0.000\n",
       "CPT_FLAG8_Sum            0.000\n",
       "DIAG_FLAG62_Sum          0.000\n",
       "PROC_FLAG2_Sum           0.000\n",
       "DIAG_FLAG27_Sum          0.000\n",
       "ndc_cat107_Sum           0.000\n",
       "ndc_cat84_Sum            0.000\n",
       "ndc_cat56_Sum            0.000\n",
       "ndc_cat57_Sum            0.000\n",
       "ndc_cat69_Sum            0.000\n",
       "REVCODE_FLAG52_Sum       0.000\n",
       "REVCODE_FLAG51_Sum       0.000\n",
       "REVCODE_FLAG50_Sum       0.000\n",
       "ndc_cat62_Sum            0.000\n",
       "ndc_cat64_Sum            0.000\n",
       "ndc_cat55_Sum            0.000\n",
       "REVCODE_FLAG47_Sum       0.000\n",
       "REVCODE_FLAG46_Sum       0.000\n",
       "ndc_cat25_Sum            0.000\n",
       "REVCODE_FLAG42_Sum       0.000\n",
       "REVCODE_FLAG58_Sum       0.000\n",
       "REVCODE_FLAG53_Sum       0.000\n",
       "ndc_cat52_Sum            0.000\n",
       "ndc_cat50_Sum            0.000\n",
       "REVCODE_FLAG54_Sum       0.000\n",
       "ndc_cat75_Sum            0.000\n",
       "REVCODE_FLAG56_Sum       0.000\n",
       "REVCODE_FLAG57_Sum       0.000\n",
       "REVCODE_FLAG21_Sum       0.000\n",
       "ndc_cat46_Sum            0.000\n",
       "REVCODE_FLAG59_Sum       0.000"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "importance = pd.DataFrame(pnu_test.feature_importances_, index=X_test.columns.values, columns=['Importance']).sort_values(by='Importance', ascending=False)\n",
    "importance.round(5) * 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x69617f98>"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAccAAAD8CAYAAAACNm9oAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x6947ff98>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ax = importance.iloc[:20].iloc[::-1].plot(kind='barh')\n",
    "ax.legend(loc='right')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Probability distribution of the 20%"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def generate_probability_table(clf, X_test, y_test):\n",
    "    probas = clf.predict_proba(X_test)[:, -1]\n",
    "    probas_df = pd.DataFrame(data={'probas':probas, 'y_test':y_test.values}).sort_values(by='probas', ascending=False)\n",
    "    bins = np.linspace(0.0, 1.0, 101)\n",
    "    percent = pd.cut(probas_df['probas'], bins=bins, include_lowest=True, precision=6, labels=list(range(0,100)))\n",
    "    probas_df['percent'] = percent\n",
    "    dummies = pd.get_dummies(probas_df['y_test'], prefix='y=', prefix_sep='')\n",
    "    probas_df = pd.concat([probas_df, dummies], axis=1)\n",
    "    probas_group = probas_df.groupby('percent')\n",
    "    percentile_df = probas_group.aggregate({'probas':'count', 'y=-1':'sum', 'y=0':'sum', 'y=1':'sum'})\n",
    "    labeled_tot = percentile_df['y=1'] + percentile_df['y=0']\n",
    "    percentile_df['unlabeled_pct'] = percentile_df['y=-1'] / percentile_df.probas\n",
    "    percentile_df['true_pos_pct'] = percentile_df['y=1'] / labeled_tot\n",
    "    percentile_df['true_neg_pct'] = percentile_df['y=0'] / labeled_tot\n",
    "    tot = percentile_df.probas.sum()\n",
    "    percentile_df['pct_of_total'] = percentile_df.probas / tot\n",
    "    percentile_df['cum_pct_of_total'] = percentile_df.pct_of_total.cumsum()\n",
    "    percentile_df = percentile_df.ffill()\n",
    "    percentile_df = percentile_df.reset_index()\n",
    "    tp = np.polyfit(x=percentile_df.index.values.astype(int), y=percentile_df.true_pos_pct, deg=1)\n",
    "    tpp = np.poly1d(tp)\n",
    "    percentile_df['tp_trendline'] = tpp(percentile_df.index.values.astype(int))\n",
    "    tn = np.polyfit(x=percentile_df.index.values.astype(int), y=percentile_df.true_neg_pct, deg=1)\n",
    "    tnp = np.poly1d(tn)\n",
    "    percentile_df['tn_trendline'] = tnp(percentile_df.index.values.astype(int))\n",
    "    return probas_df, percentile_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done  74 out of  74 | elapsed:  3.3min finished\n"
     ]
    }
   ],
   "source": [
    "probas_df, percentile_df = generate_probability_table(pnu_test, X_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x8bac1dd8>"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAY8AAAD8CAYAAACPWyg8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x4fc09f98>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "probas_df['probas'].plot.hist(bins=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "probas_df.to_csv('./res/predicted_probabilities_soft.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "df = percentile_df\n",
    "x_axis = df.index.values.astype(int)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x6a183390>"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAmoAAAFNCAYAAACwk0NsAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x72af3da0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10,5))\n",
    "plt.plot(x_axis, df.pct_of_total, 'bs', label='N')\n",
    "plt.plot(x_axis, df.cum_pct_of_total, 'b^', label='Cumulative N')\n",
    "plt.plot(x_axis, df.tp_trendline, 'g-', label='% EPI Trend')\n",
    "plt.plot(x_axis, df.tn_trendline, 'r-', label='% Non-EPI Trend')\n",
    "plt.plot(x_axis, df.true_pos_pct, 'g.', label='Labeled EPI %')\n",
    "plt.plot(x_axis, df.true_neg_pct, 'r.', label='Labeled Non-EPI %')\n",
    "plt.axis([0, 100, -0.01, 1.01])\n",
    "plt.ylabel('% of test set')\n",
    "plt.xlabel('predicted probability of EPI')\n",
    "plt.title('20% test set results')\n",
    "plt.legend(loc='right')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Precision Recall curve for various thresholds used for Model 3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.metrics import precision_recall_curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "labeled_probas = probas_df[probas_df.y_test >= 0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "pr, re, th = precision_recall_curve(labeled_probas.y_test, labeled_probas.probas, pos_label=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": true,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# find pr of unlabeled == pos\n",
    "unlabeled_probas = probas_df[probas_df.y_test == -1].probas.values\n",
    "total_unlabeled = len(unlabeled_probas)\n",
    "pr_one_un = []\n",
    "for thresh in th:\n",
    "    p = sum(unlabeled_probas >= thresh)\n",
    "    pr_one_un.append(p / total_unlabeled)\n",
    "pr_one_un.append(0.0)\n",
    "pr_one_un = np.array(pr_one_un)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def find_re_pr_prob(thresh):\n",
    "    idx = np.nonzero(th >= thresh)[0][0]\n",
    "    return re[idx], pr[idx], pr_one_un[idx]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "re_50, pr_50, _ = find_re_pr_prob(0.5)\n",
    "re_35, pr_35, _ = find_re_pr_prob(0.35)\n",
    "re_40, pr_40, _ = find_re_pr_prob(0.4)\n",
    "re_80, pr_80, _ = find_re_pr_prob(0.8)\n",
    "re_20, pr_20, _ = find_re_pr_prob(0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from matplotlib import cm, colors, colorbar"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.figure.Figure at 0x8b630160>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAlsAAAHACAYAAACVjK1sAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nMzMza6P/BzX02wE4e8LgAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x6f2af048>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.clf()\n",
    "plt.figure(figsize=(10, 7))\n",
    "plt.xlabel('Recall', fontsize=15)\n",
    "plt.xlim((0.0, 1.0))\n",
    "plt.ylabel('Precision', fontsize=15)\n",
    "plt.ylim((0.45, 1.05))\n",
    "plt.title('Precision-Recall curve of labeled 20% test set', fontsize=15)\n",
    "norm = colors.Normalize(vmin=0.0, vmax=0.2)\n",
    "cmap = cm.plasma\n",
    "plt.scatter(re, pr, c=pr_one_un, cmap=cmap, norm=norm, edgecolors='none', s=30)\n",
    "#plt.scatter(re, pr, c=np.append(th, 1.0), cmap=cm.plasma, norm=norm)\n",
    "cb = plt.colorbar(orientation='vertical', extend='max')\n",
    "cb.set_label('Probability of unlabeled positive', fontsize=15)\n",
    "plt.axvline(x=re_80, ls=\"--\", c='r', label='Threshold=80%')\n",
    "plt.axvline(x=re_50, ls=\"--\", c='g', label='Threshold=50%')\n",
    "plt.axvline(x=re_40, ls=\"--\", c='orange', label='Threshold=40%')\n",
    "plt.axvline(x=re_35, ls=\"--\", c='b', label='Threshold=35%')\n",
    "plt.axvline(x=re_20, ls=\"--\", c='y', label='Threshold=20%')\n",
    "plt.legend(loc='center left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "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": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}