notebooks/analyze-sklearn-benchmark6.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Count the number of algorithm evaluations each model had"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "classifier\n",
       "AdaBoostClassifier                4592\n",
       "BernoulliNB                      22960\n",
       "DecisionTreeClassifier          110208\n",
       "ExtraTreesClassifier            440485\n",
       "GaussianNB                         164\n",
       "GradientBoostingClassifier     2409636\n",
       "KNeighborsClassifier              8810\n",
       "LinearSVC                        18696\n",
       "LogisticRegression               39360\n",
       "MultinomialNB                     3280\n",
       "PassiveAggressiveClassifier       7202\n",
       "RandomForestClassifier          440487\n",
       "SVC                             174333\n",
       "XGBClassifier                  4380737\n",
       "Name: parameters, dtype: int64"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "data = pd.read_csv('sklearn-benchmark6-data.tsv.gz', sep='\\t', names=['dataset',\n",
    "                                                                     'classifier',\n",
    "                                                                     'parameters',\n",
    "                                                                     'accuracy', \n",
    "                                                                     'macrof1',\n",
    "                                                                     'bal_accuracy']).fillna('')\n",
    "\n",
    "data = data.groupby(['classifier', 'dataset', 'parameters'])['accuracy'].mean().reset_index()\n",
    "data.groupby('classifier')['parameters'].count()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# Rank the parameters for each model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "\n",
    "data = pd.read_csv('sklearn-benchmark6-data.tsv.gz', sep='\\t', names=['dataset',\n",
    "                                                                     'classifier',\n",
    "                                                                     'parameters',\n",
    "                                                                     'accuracy', \n",
    "                                                                     'macrof1',\n",
    "                                                                     'bal_accuracy']).fillna('')\n",
    "\n",
    "data = data.groupby(['classifier', 'dataset', 'parameters'])['accuracy'].mean().reset_index()\n",
    "data['accuracy'] = data['accuracy'].apply(lambda x: round(x, 3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[]\n",
      "number of data sets in svm_data: 164\n",
      "number of data sets: 164\n",
      "svm missing  []\n"
     ]
    }
   ],
   "source": [
    "# find data set SVM did not finish on\n",
    "svm_data = data[data['classifier']=='SVC']\n",
    "print([problem for problem,d in data.groupby('dataset') if problem not in svm_data['dataset'].unique()])\n",
    "\n",
    "print('number of data sets in svm_data:',len(svm_data['dataset'].unique()))\n",
    "print('number of data sets:',len(data['dataset'].unique()))\n",
    "print('svm missing ',[p for p in data['dataset'].unique() if p not in svm_data['dataset'].unique()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 164/164 [08:46<00:00,  3.00s/it]\n"
     ]
    }
   ],
   "source": [
    "from collections import defaultdict\n",
    "from tqdm import tqdm\n",
    "import numpy as np\n",
    "\n",
    "model_param_ranks = defaultdict(list)\n",
    "\n",
    "for dataset, group_dataset in tqdm(data.groupby('dataset')):\n",
    "    num_scores = float(len(group_dataset['accuracy'].unique()))\n",
    "    accuracy_ranks = {}\n",
    "    \n",
    "    for rank, accuracy in enumerate(sorted(group_dataset['accuracy'].unique(), reverse=True)):\n",
    "        accuracy_ranks[accuracy] = (rank + 1) / num_scores\n",
    "    \n",
    "    for index, row in group_dataset.iterrows():\n",
    "        model_param_ranks['{}-{}'.format(row['classifier'],\n",
    "                                         row['parameters'])].append(accuracy_ranks[row['accuracy']])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model_average_param_ranks = defaultdict(float)\n",
    "for model_param in model_param_ranks:\n",
    "    model_average_param_ranks[model_param] = np.mean(model_param_ranks[model_param])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Best params for each model from rankings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.081\tXGBClassifier-n_estimators=500,learning_rate=0.01,gamma=0.45,max_depth=50,subsample=0.7000000000000001,seed=324089,nthread=1\n",
      "0.092\tRandomForestClassifier-n_estimators=500,min_impurity_decrease=0.0,max_features=sqrt,criterion=entropy,random_state=324089\n",
      "0.097\tGradientBoostingClassifier-n_estimators=500,min_impurity_decrease=0.005,max_features=sqrt,learning_rate=0.1,loss=deviance,random_state=324089\n",
      "0.098\tExtraTreesClassifier-n_estimators=500,min_impurity_decrease=0.0,max_features=0.5,criterion=gini,random_state=324089\n",
      "0.127\tSVC-C=10.0,gamma=auto,kernel=poly,degree=2,coef0=0.5,random_state=324089\n",
      "0.163\tKNeighborsClassifier-n_neighbors=14,weights=distance\n",
      "0.172\tDecisionTreeClassifier-min_impurity_decrease=0.0,max_features=None,criterion=entropy,random_state=324089\n",
      "0.184\tLogisticRegression-C=2.0,penalty=l1,fit_intercept=True,dual=False,random_state=324089\n",
      "0.188\tLinearSVC-C=1.0,penalty=l1,fit_intercept=True,dual=False,random_state=324089\n",
      "0.222\tPassiveAggressiveClassifier-C=0.01,loss=squared_hinge,fit_intercept=True,random_state=324089\n",
      "0.225\tAdaBoostClassifier-learning_rate=0.5,n_estimators=50,random_state=324089\n",
      "0.314\tBernoulliNB-alpha=0.1,fit_prior=True,binarize=0.5\n",
      "0.331\tGaussianNB-\n",
      "0.343\tMultinomialNB-alpha=0.0,fit_prior=True\n"
     ]
    }
   ],
   "source": [
    "models_seen = set()\n",
    "\n",
    "for model_param in sorted(model_average_param_ranks, key=model_average_param_ranks.get, reverse=False):\n",
    "    model = model_param.split('-')[0]\n",
    "    if model not in models_seen:\n",
    "        models_seen.add(model)\n",
    "    else:\n",
    "        continue\n",
    "\n",
    "    print('{}\\t{}'.format(round(model_average_param_ranks[model_param], 3), model_param))\n",
    "    \n",
    "    if len(models_seen) >= 15:\n",
    "        break"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Average each model parameter's rankings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from collections import defaultdict\n",
    "import numpy as np\n",
    "\n",
    "model_param_breakdown_rankings = defaultdict(lambda: defaultdict(lambda: defaultdict(list)))\n",
    "\n",
    "for model_param in model_average_param_ranks:\n",
    "    model = model_param.split('-')[0]\n",
    "    if model == 'GaussianNB':\n",
    "        continue\n",
    "    params = '-'.join(model_param.split('-')[1:])\n",
    "    params = params.split(',')\n",
    "    rank = model_average_param_ranks[model_param]\n",
    "    for param in params:\n",
    "        model_param_breakdown_rankings[model][param.split('=')[0]][param.split('=')[-1]].append(rank)\n",
    "\n",
    "model_param_breakdown_average_rankings = defaultdict(lambda: defaultdict(lambda: defaultdict(float)))\n",
    "\n",
    "for model in sorted(model_param_breakdown_rankings):\n",
    "    for param in model_param_breakdown_rankings[model]:\n",
    "        for param_val in model_param_breakdown_rankings[model][param]:\n",
    "            model_param_breakdown_average_rankings[model][param][param_val] = round(np.mean(\n",
    "                model_param_breakdown_rankings[model][param][param_val]), 3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AdaBoostClassifier\n",
      "--------------------\n",
      "learning_rate\n",
      "\t=0.5                      0.233\n",
      "\t=0.1                      0.246\n",
      "\t=1.0                      0.262\n",
      "\t=0.01                     0.287\n",
      "\t=50.0                     0.679\n",
      "\t=100.0                    0.709\n",
      "\t=10.0                     0.797\n",
      "n_estimators\n",
      "\t=500                      0.447\n",
      "\t=100                      0.453\n",
      "\t=50                       0.458\n",
      "\t=10                       0.478\n",
      "random_state\n",
      "\t=324089                   0.459\n",
      "\n",
      "BernoulliNB\n",
      "--------------------\n",
      "alpha\n",
      "\t=10.0                     0.42\n",
      "\t=5.0                      0.422\n",
      "\t=25.0                     0.424\n",
      "\t=1.0                      0.429\n",
      "\t=0.75                     0.43\n",
      "\t=0.5                      0.431\n",
      "\t=50.0                     0.431\n",
      "\t=0.25                     0.432\n",
      "\t=0.1                      0.434\n",
      "\t=0.0                      0.438\n",
      "fit_prior\n",
      "\t=True                     0.401\n",
      "\t=False                    0.457\n",
      "binarize\n",
      "\t=0.5                      0.351\n",
      "\t=0.25                     0.363\n",
      "\t=0.75                     0.394\n",
      "\t=0.1                      0.401\n",
      "\t=0.9                      0.442\n",
      "\t=0.0                      0.46\n",
      "\t=1.0                      0.593\n",
      "\n",
      "DecisionTreeClassifier\n",
      "--------------------\n",
      "min_impurity_decrease\n",
      "\t=0.0                      0.207\n",
      "\t=0.00025                  0.221\n",
      "\t=0.0005                   0.225\n",
      "\t=0.00075                  0.228\n",
      "\t=0.001                    0.231\n",
      "\t=0.00125                  0.232\n",
      "\t=0.0015                   0.235\n",
      "\t=0.00175                  0.237\n",
      "\t=0.002                    0.238\n",
      "\t=0.0022500000000000003    0.24\n",
      "\t=0.0025                   0.241\n",
      "\t=0.00275                  0.243\n",
      "\t=0.003                    0.244\n",
      "\t=0.0032500000000000003    0.246\n",
      "\t=0.0035                   0.247\n",
      "\t=0.00375                  0.248\n",
      "\t=0.004                    0.25\n",
      "\t=0.00425                  0.25\n",
      "\t=0.0045000000000000005    0.251\n",
      "\t=0.00475                  0.253\n",
      "\t=0.005                    0.254\n",
      "\t=0.0075                   0.264\n",
      "\t=0.01                     0.273\n",
      "\t=0.0125                   0.282\n",
      "\t=0.015                    0.288\n",
      "\t=0.0175                   0.295\n",
      "\t=0.02                     0.302\n",
      "\t=0.0225                   0.308\n",
      "\t=0.025                    0.314\n",
      "\t=0.0275                   0.32\n",
      "\t=0.03                     0.325\n",
      "\t=0.0325                   0.331\n",
      "\t=0.035                    0.336\n",
      "\t=0.0375                   0.34\n",
      "\t=0.04                     0.345\n",
      "\t=0.0425                   0.349\n",
      "\t=0.045                    0.355\n",
      "\t=0.0475                   0.358\n",
      "\t=0.05                     0.362\n",
      "\t=0.1                      0.417\n",
      "\t=0.15000000000000002      0.454\n",
      "\t=0.2                      0.481\n",
      "\t=0.25                     0.501\n",
      "\t=0.30000000000000004      0.52\n",
      "\t=0.35000000000000003      0.537\n",
      "\t=0.4                      0.545\n",
      "\t=0.45                     0.551\n",
      "\t=0.5                      0.558\n",
      "max_features\n",
      "\t=None                     0.257\n",
      "\t=0.75                     0.272\n",
      "\t=0.5                      0.29\n",
      "\t=0.25                     0.341\n",
      "\t=sqrt                     0.343\n",
      "\t=log2                     0.347\n",
      "\t=0.1                      0.387\n",
      "criterion\n",
      "\t=entropy                  0.301\n",
      "\t=gini                     0.338\n",
      "random_state\n",
      "\t=324089                   0.319\n",
      "\n",
      "ExtraTreesClassifier\n",
      "--------------------\n",
      "n_estimators\n",
      "\t=500                      0.261\n",
      "\t=100                      0.262\n",
      "\t=50                       0.264\n",
      "\t=10                       0.282\n",
      "min_impurity_decrease\n",
      "\t=0.0                      0.113\n",
      "\t=0.00025                  0.129\n",
      "\t=0.0005                   0.139\n",
      "\t=0.00075                  0.146\n",
      "\t=0.001                    0.152\n",
      "\t=0.00125                  0.157\n",
      "\t=0.0015                   0.161\n",
      "\t=0.00175                  0.165\n",
      "\t=0.002                    0.168\n",
      "\t=0.0022500000000000003    0.17\n",
      "\t=0.0025                   0.173\n",
      "\t=0.00275                  0.175\n",
      "\t=0.003                    0.178\n",
      "\t=0.0032500000000000003    0.18\n",
      "\t=0.0035                   0.182\n",
      "\t=0.00375                  0.184\n",
      "\t=0.004                    0.186\n",
      "\t=0.00425                  0.187\n",
      "\t=0.0045000000000000005    0.188\n",
      "\t=0.00475                  0.19\n",
      "\t=0.005                    0.192\n",
      "\t=0.0075                   0.206\n",
      "\t=0.01                     0.218\n",
      "\t=0.0125                   0.227\n",
      "\t=0.015                    0.235\n",
      "\t=0.0175                   0.244\n",
      "\t=0.02                     0.252\n",
      "\t=0.0225                   0.259\n",
      "\t=0.025                    0.265\n",
      "\t=0.0275                   0.271\n",
      "\t=0.03                     0.277\n",
      "\t=0.0325                   0.283\n",
      "\t=0.035                    0.289\n",
      "\t=0.0375                   0.294\n",
      "\t=0.04                     0.299\n",
      "\t=0.0425                   0.304\n",
      "\t=0.045                    0.308\n",
      "\t=0.0475                   0.313\n",
      "\t=0.05                     0.317\n",
      "\t=0.1                      0.396\n",
      "\t=0.15000000000000002      0.434\n",
      "\t=0.2                      0.463\n",
      "\t=0.25                     0.486\n",
      "\t=0.30000000000000004      0.508\n",
      "\t=0.35000000000000003      0.521\n",
      "\t=0.4                      0.532\n",
      "\t=0.45                     0.546\n",
      "\t=0.5                      0.555\n",
      "max_features\n",
      "\t=None                     0.232\n",
      "\t=0.75                     0.237\n",
      "\t=0.5                      0.247\n",
      "\t=sqrt                     0.275\n",
      "\t=log2                     0.278\n",
      "\t=0.25                     0.279\n",
      "\t=0.1                      0.322\n",
      "criterion\n",
      "\t=entropy                  0.241\n",
      "\t=gini                     0.293\n",
      "random_state\n",
      "\t=324089                   0.267\n",
      "\n",
      "GradientBoostingClassifier\n",
      "--------------------\n",
      "n_estimators\n",
      "\t=500                      0.319\n",
      "\t=100                      0.338\n",
      "\t=50                       0.354\n",
      "\t=10                       0.413\n",
      "min_impurity_decrease\n",
      "\t=0.00425                  0.351\n",
      "\t=0.0045000000000000005    0.351\n",
      "\t=0.00475                  0.351\n",
      "\t=0.00025                  0.352\n",
      "\t=0.0005                   0.352\n",
      "\t=0.00075                  0.352\n",
      "\t=0.001                    0.352\n",
      "\t=0.00125                  0.352\n",
      "\t=0.0015                   0.352\n",
      "\t=0.00175                  0.352\n",
      "\t=0.002                    0.352\n",
      "\t=0.0022500000000000003    0.352\n",
      "\t=0.00275                  0.352\n",
      "\t=0.003                    0.352\n",
      "\t=0.0032500000000000003    0.352\n",
      "\t=0.0035                   0.352\n",
      "\t=0.00375                  0.352\n",
      "\t=0.004                    0.352\n",
      "\t=0.0                      0.353\n",
      "\t=0.0025                   0.353\n",
      "\t=0.005                    0.353\n",
      "\t=0.0075                   0.353\n",
      "\t=0.01                     0.353\n",
      "\t=0.0125                   0.354\n",
      "\t=0.015                    0.354\n",
      "\t=0.0175                   0.354\n",
      "\t=0.02                     0.354\n",
      "\t=0.0225                   0.354\n",
      "\t=0.025                    0.354\n",
      "\t=0.0275                   0.355\n",
      "\t=0.03                     0.355\n",
      "\t=0.0325                   0.355\n",
      "\t=0.035                    0.355\n",
      "\t=0.0375                   0.355\n",
      "\t=0.04                     0.356\n",
      "\t=0.0425                   0.356\n",
      "\t=0.045                    0.356\n",
      "\t=0.0475                   0.356\n",
      "\t=0.05                     0.358\n",
      "\t=0.1                      0.362\n",
      "\t=0.15000000000000002      0.363\n",
      "\t=0.2                      0.365\n",
      "\t=0.25                     0.367\n",
      "\t=0.30000000000000004      0.369\n",
      "\t=0.35000000000000003      0.37\n",
      "\t=0.4                      0.371\n",
      "\t=0.45                     0.372\n",
      "\t=0.5                      0.373\n",
      "max_features\n",
      "\t=0.5                      0.347\n",
      "\t=0.25                     0.348\n",
      "\t=sqrt                     0.352\n",
      "\t=log2                     0.356\n",
      "\t=0.1                      0.357\n",
      "\t=0.75                     0.357\n",
      "\t=None                     0.377\n",
      "learning_rate\n",
      "\t=0.5                      0.142\n",
      "\t=0.1                      0.146\n",
      "\t=1.0                      0.186\n",
      "\t=0.01                     0.242\n",
      "\t=10.0                     0.56\n",
      "\t=50.0                     0.603\n",
      "\t=100.0                    0.615\n",
      "loss\n",
      "\t=exponential              0.295\n",
      "\t=deviance                 0.417\n",
      "random_state\n",
      "\t=324089                   0.356\n",
      "\n",
      "KNeighborsClassifier\n",
      "--------------------\n",
      "n_neighbors\n",
      "\t=14                       0.178\n",
      "\t=12                       0.179\n",
      "\t=7                        0.179\n",
      "\t=10                       0.18\n",
      "\t=13                       0.18\n",
      "\t=11                       0.181\n",
      "\t=15                       0.181\n",
      "\t=16                       0.182\n",
      "\t=8                        0.182\n",
      "\t=9                        0.183\n",
      "\t=6                        0.184\n",
      "\t=17                       0.186\n",
      "\t=5                        0.186\n",
      "\t=18                       0.187\n",
      "\t=19                       0.193\n",
      "\t=4                        0.193\n",
      "\t=20                       0.195\n",
      "\t=3                        0.195\n",
      "\t=21                       0.199\n",
      "\t=22                       0.201\n",
      "\t=23                       0.203\n",
      "\t=24                       0.203\n",
      "\t=25                       0.205\n",
      "\t=1                        0.218\n",
      "\t=50                       0.221\n",
      "\t=2                        0.226\n",
      "\t=100                      0.24\n",
      "weights\n",
      "\t=distance                 0.181\n",
      "\t=uniform                  0.207\n",
      "\n",
      "LinearSVC\n",
      "--------------------\n",
      "C\n",
      "\t=0.5                      0.222\n",
      "\t=0.6000000000000001       0.222\n",
      "\t=0.7000000000000001       0.222\n",
      "\t=0.8                      0.222\n",
      "\t=0.9                      0.222\n",
      "\t=1.0                      0.222\n",
      "\t=0.4                      0.223\n",
      "\t=2.0                      0.223\n",
      "\t=3.0                      0.223\n",
      "\t=0.2                      0.224\n",
      "\t=0.30000000000000004      0.224\n",
      "\t=10.0                     0.224\n",
      "\t=4.0                      0.224\n",
      "\t=5.0                      0.224\n",
      "\t=6.0                      0.224\n",
      "\t=7.0                      0.224\n",
      "\t=8.0                      0.224\n",
      "\t=9.0                      0.224\n",
      "\t=0.1                      0.228\n",
      "penalty\n",
      "\t=l1                       0.223\n",
      "\t=l2                       0.224\n",
      "fit_intercept\n",
      "\t=True                     0.193\n",
      "\t=False                    0.254\n",
      "dual\n",
      "\t=False                    0.223\n",
      "\t=True                     0.223\n",
      "random_state\n",
      "\t=324089                   0.223\n",
      "\n",
      "LogisticRegression\n",
      "--------------------\n",
      "C\n",
      "\t=5.0                      0.219\n",
      "\t=5.5                      0.219\n",
      "\t=6.0                      0.219\n",
      "\t=6.5                      0.219\n",
      "\t=7.0                      0.219\n",
      "\t=7.5                      0.219\n",
      "\t=10.0                     0.22\n",
      "\t=10.5                     0.22\n",
      "\t=11.0                     0.22\n",
      "\t=11.5                     0.22\n",
      "\t=12.0                     0.22\n",
      "\t=12.5                     0.22\n",
      "\t=13.0                     0.22\n",
      "\t=13.5                     0.22\n",
      "\t=14.0                     0.22\n",
      "\t=14.5                     0.22\n",
      "\t=15.0                     0.22\n",
      "\t=15.5                     0.22\n",
      "\t=16.0                     0.22\n",
      "\t=16.5                     0.22\n",
      "\t=17.0                     0.22\n",
      "\t=17.5                     0.22\n",
      "\t=2.0                      0.22\n",
      "\t=2.5                      0.22\n",
      "\t=3.0                      0.22\n",
      "\t=3.5                      0.22\n",
      "\t=4.0                      0.22\n",
      "\t=4.5                      0.22\n",
      "\t=8.0                      0.22\n",
      "\t=8.5                      0.22\n",
      "\t=9.0                      0.22\n",
      "\t=9.5                      0.22\n",
      "\t=1.5                      0.221\n",
      "\t=18.0                     0.221\n",
      "\t=18.5                     0.221\n",
      "\t=19.0                     0.221\n",
      "\t=19.5                     0.221\n",
      "\t=20.0                     0.221\n",
      "\t=1.0                      0.222\n",
      "\t=0.5                      0.226\n",
      "penalty\n",
      "\t=l1                       0.218\n",
      "\t=l2                       0.221\n",
      "fit_intercept\n",
      "\t=True                     0.188\n",
      "\t=False                    0.252\n",
      "dual\n",
      "\t=False                    0.219\n",
      "\t=True                     0.222\n",
      "random_state\n",
      "\t=324089                   0.22\n",
      "\n",
      "MultinomialNB\n",
      "--------------------\n",
      "alpha\n",
      "\t=0.75                     0.368\n",
      "\t=0.25                     0.369\n",
      "\t=0.5                      0.369\n",
      "\t=1.0                      0.369\n",
      "\t=5.0                      0.369\n",
      "\t=0.0                      0.37\n",
      "\t=0.1                      0.37\n",
      "\t=10.0                     0.373\n",
      "\t=25.0                     0.382\n",
      "\t=50.0                     0.392\n",
      "fit_prior\n",
      "\t=True                     0.354\n",
      "\t=False                    0.393\n",
      "\n",
      "PassiveAggressiveClassifier\n",
      "--------------------\n",
      "C\n",
      "\t=0.01                     0.252\n",
      "\t=0.001                    0.277\n",
      "\t=0.1                      0.286\n",
      "\t=0.0001                   0.324\n",
      "\t=0.5                      0.325\n",
      "\t=1.0                      0.338\n",
      "\t=10.0                     0.353\n",
      "\t=100.0                    0.355\n",
      "\t=50.0                     0.355\n",
      "\t=1e-05                    0.366\n",
      "\t=1e-06                    0.384\n",
      "loss\n",
      "\t=squared_hinge            0.326\n",
      "\t=hinge                    0.331\n",
      "fit_intercept\n",
      "\t=True                     0.303\n",
      "\t=False                    0.354\n",
      "random_state\n",
      "\t=324089                   0.329\n",
      "\n",
      "RandomForestClassifier\n",
      "--------------------\n",
      "n_estimators\n",
      "\t=500                      0.217\n",
      "\t=100                      0.219\n",
      "\t=50                       0.221\n",
      "\t=10                       0.237\n",
      "min_impurity_decrease\n",
      "\t=0.0                      0.108\n",
      "\t=0.00025                  0.113\n",
      "\t=0.0005                   0.116\n",
      "\t=0.00075                  0.119\n",
      "\t=0.001                    0.122\n",
      "\t=0.00125                  0.125\n",
      "\t=0.0015                   0.127\n",
      "\t=0.00175                  0.13\n",
      "\t=0.002                    0.132\n",
      "\t=0.0022500000000000003    0.134\n",
      "\t=0.0025                   0.135\n",
      "\t=0.00275                  0.137\n",
      "\t=0.003                    0.139\n",
      "\t=0.0032500000000000003    0.141\n",
      "\t=0.0035                   0.143\n",
      "\t=0.00375                  0.145\n",
      "\t=0.004                    0.146\n",
      "\t=0.00425                  0.148\n",
      "\t=0.0045000000000000005    0.149\n",
      "\t=0.00475                  0.15\n",
      "\t=0.005                    0.152\n",
      "\t=0.0075                   0.162\n",
      "\t=0.01                     0.17\n",
      "\t=0.0125                   0.177\n",
      "\t=0.015                    0.184\n",
      "\t=0.0175                   0.19\n",
      "\t=0.02                     0.196\n",
      "\t=0.0225                   0.202\n",
      "\t=0.025                    0.208\n",
      "\t=0.0275                   0.213\n",
      "\t=0.03                     0.219\n",
      "\t=0.0325                   0.223\n",
      "\t=0.035                    0.229\n",
      "\t=0.0375                   0.234\n",
      "\t=0.04                     0.239\n",
      "\t=0.0425                   0.244\n",
      "\t=0.045                    0.248\n",
      "\t=0.0475                   0.253\n",
      "\t=0.05                     0.256\n",
      "\t=0.1                      0.334\n",
      "\t=0.15000000000000002      0.379\n",
      "\t=0.2                      0.412\n",
      "\t=0.25                     0.441\n",
      "\t=0.30000000000000004      0.467\n",
      "\t=0.35000000000000003      0.489\n",
      "\t=0.4                      0.505\n",
      "\t=0.45                     0.517\n",
      "\t=0.5                      0.531\n",
      "max_features\n",
      "\t=0.75                     0.209\n",
      "\t=None                     0.21\n",
      "\t=0.5                      0.212\n",
      "\t=sqrt                     0.225\n",
      "\t=log2                     0.227\n",
      "\t=0.25                     0.228\n",
      "\t=0.1                      0.254\n",
      "criterion\n",
      "\t=entropy                  0.199\n",
      "\t=gini                     0.248\n",
      "random_state\n",
      "\t=324089                   0.224\n",
      "\n",
      "SVC\n",
      "--------------------\n",
      "C\n",
      "\t=100.0                    0.288\n",
      "\t=50.0                     0.288\n",
      "\t=10.0                     0.292\n",
      "\t=1.0                      0.303\n",
      "\t=0.5                      0.308\n",
      "\t=0.1                      0.32\n",
      "\t=0.01                     0.343\n",
      "gamma\n",
      "\t=auto                     0.29\n",
      "\t=0.1                      0.292\n",
      "\t=0.5                      0.297\n",
      "\t=1.0                      0.301\n",
      "\t=10.0                     0.313\n",
      "\t=50.0                     0.315\n",
      "\t=100.0                    0.316\n",
      "\t=0.01                     0.325\n",
      "kernel\n",
      "\t=poly                     0.192\n",
      "\t=rbf                      0.34\n",
      "\t=sigmoid                  0.53\n",
      "degree\n",
      "\t=3                        0.194\n",
      "\t=2                        0.359\n",
      "coef0\n",
      "\t=1.0                      0.293\n",
      "\t=0.5                      0.296\n",
      "\t=10.0                     0.298\n",
      "\t=50.0                     0.304\n",
      "\t=100.0                    0.305\n",
      "\t=0.1                      0.307\n",
      "\t=0.0                      0.332\n",
      "random_state\n",
      "\t=324089                   0.306\n",
      "\n",
      "XGBClassifier\n",
      "--------------------\n",
      "n_estimators\n",
      "\t=500                      0.327\n",
      "\t=100                      0.334\n",
      "\t=50                       0.338\n",
      "\t=10                       0.348\n",
      "learning_rate\n",
      "\t=0.1                      0.181\n",
      "\t=0.5                      0.182\n",
      "\t=1.0                      0.198\n",
      "\t=0.01                     0.206\n",
      "\t=100.0                    0.49\n",
      "\t=50.0                     0.519\n",
      "\t=10.0                     0.58\n",
      "gamma\n",
      "\t=0.45                     0.335\n",
      "\t=0.5                      0.335\n",
      "\t=0.30000000000000004      0.336\n",
      "\t=0.35000000000000003      0.336\n",
      "\t=0.4                      0.336\n",
      "\t=0.2                      0.337\n",
      "\t=0.25                     0.337\n",
      "\t=0.0                      0.338\n",
      "\t=0.05                     0.338\n",
      "\t=0.1                      0.338\n",
      "\t=0.15000000000000002      0.338\n",
      "max_depth\n",
      "\t=3                        0.331\n",
      "\t=20                       0.332\n",
      "\t=2                        0.333\n",
      "\t=50                       0.333\n",
      "\t=10                       0.336\n",
      "\t=4                        0.339\n",
      "\t=5                        0.34\n",
      "\t=1                        0.351\n",
      "subsample\n",
      "\t=1.0                      0.283\n",
      "\t=0.9                      0.286\n",
      "\t=0.7000000000000001       0.288\n",
      "\t=0.8                      0.289\n",
      "\t=0.6000000000000001       0.292\n",
      "\t=0.5                      0.294\n",
      "\t=0.4                      0.297\n",
      "\t=0.30000000000000004      0.3\n",
      "\t=0.2                      0.307\n",
      "\t=0.1                      0.328\n",
      "\t=0.0                      0.74\n",
      "seed\n",
      "\t=324089                   0.337\n",
      "nthread\n",
      "\t=1                        0.337\n",
      "\n"
     ]
    }
   ],
   "source": [
    "for model in sorted(model_param_breakdown_average_rankings):\n",
    "    print(model)\n",
    "    print('--------------------')\n",
    "    for param in model_param_breakdown_average_rankings[model]:\n",
    "        print(param)\n",
    "        for param_val in sorted(model_param_breakdown_average_rankings[model][param],\n",
    "                                key=model_param_breakdown_average_rankings[model][param].get):\n",
    "            print('\\t={}{}{}'.format(param_val,\n",
    "                                   (' ' * 25)[:25 - len(param_val)],\n",
    "                                   model_param_breakdown_average_rankings[model][param][param_val]))\n",
    "            \n",
    "    print('')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Rank each model on a per-data set basis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "\n",
    "data = pd.read_csv('sklearn-benchmark6-data.tsv.gz', sep='\\t', names=['dataset',\n",
    "                                                                     'classifier',\n",
    "                                                                     'parameters',\n",
    "                                                                     'accuracy', \n",
    "                                                                     'macrof1',\n",
    "                                                                     'bal_accuracy']).fillna('')\n",
    "\n",
    "data = data.groupby(['classifier', 'dataset', 'parameters'])['accuracy'].mean().reset_index()\n",
    "data['accuracy'] = data['accuracy'].apply(lambda x: round(x, 3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 164/164 [07:13<00:00,  2.54s/it]\n"
     ]
    }
   ],
   "source": [
    "from collections import defaultdict\n",
    "from tqdm import tqdm\n",
    "import numpy as np\n",
    "\n",
    "model_ranks = defaultdict(list)\n",
    "\n",
    "for dataset, group_dataset in tqdm(data.groupby('dataset')):\n",
    "    if len(group_dataset['classifier'].unique()) != 14:\n",
    "        continue\n",
    "    \n",
    "    num_scores = float(len(group_dataset['accuracy'].unique()))\n",
    "    accuracy_ranks = {}\n",
    "    \n",
    "    for rank, accuracy in enumerate(sorted(group_dataset['accuracy'].unique(), reverse=True)):\n",
    "        accuracy_ranks[accuracy] = (rank + 1) / num_scores\n",
    "    \n",
    "    for index, row in group_dataset.iterrows():\n",
    "        model_ranks[row['classifier']].append(accuracy_ranks[row['accuracy']])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model_average_ranks = defaultdict(float)\n",
    "for model in model_ranks:\n",
    "    model_average_ranks[model] = np.mean(model_ranks[model])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.194\tKNeighborsClassifier\n",
      "0.22\tLogisticRegression\n",
      "0.223\tLinearSVC\n",
      "0.224\tRandomForestClassifier\n",
      "0.267\tExtraTreesClassifier\n",
      "0.319\tDecisionTreeClassifier\n",
      "0.325\tSVC\n",
      "0.329\tPassiveAggressiveClassifier\n",
      "0.331\tGaussianNB\n",
      "0.336\tXGBClassifier\n",
      "0.372\tGradientBoostingClassifier\n",
      "0.373\tMultinomialNB\n",
      "0.429\tBernoulliNB\n",
      "0.459\tAdaBoostClassifier\n"
     ]
    }
   ],
   "source": [
    "for model in sorted(model_average_ranks, key=model_average_ranks.get, reverse=False):\n",
    "    print('{}\\t{}'.format(round(model_average_ranks[model], 3), model))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# How often is model X better than model Y?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "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>dataset</th>\n",
       "      <th>classifier</th>\n",
       "      <th>accuracy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...</td>\n",
       "      <td>AdaBoostClassifier</td>\n",
       "      <td>0.504</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...</td>\n",
       "      <td>BernoulliNB</td>\n",
       "      <td>0.512</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...</td>\n",
       "      <td>DecisionTreeClassifier</td>\n",
       "      <td>0.527</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...</td>\n",
       "      <td>ExtraTreesClassifier</td>\n",
       "      <td>0.536</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...</td>\n",
       "      <td>GaussianNB</td>\n",
       "      <td>0.501</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                             dataset              classifier  \\\n",
       "0  GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...      AdaBoostClassifier   \n",
       "1  GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...             BernoulliNB   \n",
       "2  GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...  DecisionTreeClassifier   \n",
       "3  GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...    ExtraTreesClassifier   \n",
       "4  GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...              GaussianNB   \n",
       "\n",
       "   accuracy  \n",
       "0     0.504  \n",
       "1     0.512  \n",
       "2     0.527  \n",
       "3     0.536  \n",
       "4     0.501  "
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import pdb\n",
    "\n",
    "data = pd.read_csv('sklearn-benchmark6-data.tsv.gz', sep='\\t', names=['dataset',\n",
    "                                                                     'classifier',\n",
    "                                                                     'parameters',\n",
    "                                                                     'accuracy', \n",
    "                                                                     'macrof1',\n",
    "                                                                     'bal_accuracy']).fillna('')\n",
    "\n",
    "data = data.groupby(['dataset','classifier'])['accuracy'].max().reset_index()\n",
    "data['accuracy'] = data['accuracy'].apply(lambda x: round(x, 3))\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/164 [00:00<?, ?it/s]/Users/randal_olson/anaconda/lib/python3.6/site-packages/pandas/core/indexing.py:517: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  self.obj[item] = s\n",
      "100%|██████████| 164/164 [00:28<00:00,  5.60it/s]\n"
     ]
    }
   ],
   "source": [
    "from collections import defaultdict\n",
    "from tqdm import tqdm\n",
    "import numpy as np\n",
    "from itertools import combinations\n",
    "import pdb\n",
    "\n",
    "model_tourneys = defaultdict(int)\n",
    "all_models = sorted(data['classifier'].unique())\n",
    "\n",
    "for dataset, group_dataset in tqdm(data.groupby('dataset')):\n",
    "    \n",
    "#     pdb.set_trace()\n",
    "    group_dataset.loc[:, 'accuracy']= group_dataset['accuracy'].values / group_dataset['accuracy'].max()\n",
    "    group_dataset = group_dataset.set_index('classifier')['accuracy'].to_dict()\n",
    "\n",
    "    for (model1, model2) in combinations(group_dataset.keys(), 2):\n",
    "        if group_dataset[model1] >= group_dataset[model2] + 0.01:\n",
    "            model_tourneys[(model1, model2)] += 1\n",
    "        elif group_dataset[model2] >= group_dataset[model1] + 0.01:\n",
    "            model_tourneys[(model2, model1)] += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AdaBoostClassifier                    W / L\n",
      "--------------------\n",
      "\tBernoulliNB                   98 / 27\n",
      "\tDecisionTreeClassifier        35 / 81\n",
      "\tExtraTreesClassifier          11 / 116\n",
      "\tGaussianNB                    131 / 16\n",
      "\tGradientBoostingClassifier    3 / 129\n",
      "\tKNeighborsClassifier          55 / 72\n",
      "\tLinearSVC                     60 / 52\n",
      "\tLogisticRegression            58 / 58\n",
      "\tMultinomialNB                 126 / 22\n",
      "\tPassiveAggressiveClassifier   73 / 51\n",
      "\tRandomForestClassifier        9 / 118\n",
      "\tSVC                           15 / 101\n",
      "\tXGBClassifier                 0 / 132\n",
      "\n",
      "BernoulliNB                           W / L\n",
      "--------------------\n",
      "\tAdaBoostClassifier            27 / 98\n",
      "\tDecisionTreeClassifier        17 / 119\n",
      "\tExtraTreesClassifier          1 / 146\n",
      "\tGaussianNB                    99 / 39\n",
      "\tGradientBoostingClassifier    1 / 149\n",
      "\tKNeighborsClassifier          29 / 119\n",
      "\tLinearSVC                     28 / 88\n",
      "\tLogisticRegression            24 / 93\n",
      "\tMultinomialNB                 107 / 25\n",
      "\tPassiveAggressiveClassifier   38 / 85\n",
      "\tRandomForestClassifier        1 / 143\n",
      "\tSVC                           5 / 138\n",
      "\tXGBClassifier                 0 / 151\n",
      "\n",
      "DecisionTreeClassifier                W / L\n",
      "--------------------\n",
      "\tAdaBoostClassifier            81 / 35\n",
      "\tBernoulliNB                   119 / 17\n",
      "\tExtraTreesClassifier          4 / 107\n",
      "\tGaussianNB                    137 / 15\n",
      "\tGradientBoostingClassifier    2 / 116\n",
      "\tKNeighborsClassifier          72 / 49\n",
      "\tLinearSVC                     87 / 38\n",
      "\tLogisticRegression            85 / 42\n",
      "\tMultinomialNB                 137 / 10\n",
      "\tPassiveAggressiveClassifier   99 / 35\n",
      "\tRandomForestClassifier        0 / 105\n",
      "\tSVC                           26 / 83\n",
      "\tXGBClassifier                 1 / 124\n",
      "\n",
      "ExtraTreesClassifier                  W / L\n",
      "--------------------\n",
      "\tAdaBoostClassifier            116 / 11\n",
      "\tBernoulliNB                   146 / 1\n",
      "\tDecisionTreeClassifier        107 / 4\n",
      "\tGaussianNB                    159 / 0\n",
      "\tGradientBoostingClassifier    28 / 40\n",
      "\tKNeighborsClassifier          128 / 3\n",
      "\tLinearSVC                     121 / 8\n",
      "\tLogisticRegression            116 / 10\n",
      "\tMultinomialNB                 157 / 2\n",
      "\tPassiveAggressiveClassifier   133 / 5\n",
      "\tRandomForestClassifier        27 / 28\n",
      "\tSVC                           59 / 27\n",
      "\tXGBClassifier                 22 / 51\n",
      "\n",
      "GaussianNB                            W / L\n",
      "--------------------\n",
      "\tAdaBoostClassifier            16 / 131\n",
      "\tBernoulliNB                   39 / 99\n",
      "\tDecisionTreeClassifier        15 / 137\n",
      "\tExtraTreesClassifier          0 / 159\n",
      "\tGradientBoostingClassifier    0 / 159\n",
      "\tKNeighborsClassifier          7 / 140\n",
      "\tLinearSVC                     14 / 132\n",
      "\tLogisticRegression            14 / 132\n",
      "\tMultinomialNB                 77 / 66\n",
      "\tPassiveAggressiveClassifier   18 / 129\n",
      "\tRandomForestClassifier        1 / 158\n",
      "\tSVC                           5 / 154\n",
      "\tXGBClassifier                 1 / 158\n",
      "\n",
      "GradientBoostingClassifier            W / L\n",
      "--------------------\n",
      "\tAdaBoostClassifier            129 / 3\n",
      "\tBernoulliNB                   149 / 1\n",
      "\tDecisionTreeClassifier        116 / 2\n",
      "\tExtraTreesClassifier          40 / 28\n",
      "\tGaussianNB                    159 / 0\n",
      "\tKNeighborsClassifier          115 / 9\n",
      "\tLinearSVC                     124 / 6\n",
      "\tLogisticRegression            122 / 7\n",
      "\tMultinomialNB                 156 / 2\n",
      "\tPassiveAggressiveClassifier   137 / 3\n",
      "\tRandomForestClassifier        37 / 20\n",
      "\tSVC                           67 / 24\n",
      "\tXGBClassifier                 12 / 34\n",
      "\n",
      "KNeighborsClassifier                  W / L\n",
      "--------------------\n",
      "\tAdaBoostClassifier            72 / 55\n",
      "\tBernoulliNB                   119 / 29\n",
      "\tDecisionTreeClassifier        49 / 72\n",
      "\tExtraTreesClassifier          3 / 128\n",
      "\tGaussianNB                    140 / 7\n",
      "\tGradientBoostingClassifier    9 / 115\n",
      "\tLinearSVC                     72 / 47\n",
      "\tLogisticRegression            70 / 55\n",
      "\tMultinomialNB                 137 / 10\n",
      "\tPassiveAggressiveClassifier   90 / 40\n",
      "\tRandomForestClassifier        4 / 120\n",
      "\tSVC                           8 / 111\n",
      "\tXGBClassifier                 4 / 130\n",
      "\n",
      "LinearSVC                             W / L\n",
      "--------------------\n",
      "\tAdaBoostClassifier            52 / 60\n",
      "\tBernoulliNB                   88 / 28\n",
      "\tDecisionTreeClassifier        38 / 87\n",
      "\tExtraTreesClassifier          8 / 121\n",
      "\tGaussianNB                    132 / 14\n",
      "\tGradientBoostingClassifier    6 / 124\n",
      "\tKNeighborsClassifier          47 / 72\n",
      "\tLogisticRegression            6 / 25\n",
      "\tMultinomialNB                 137 / 3\n",
      "\tPassiveAggressiveClassifier   68 / 19\n",
      "\tRandomForestClassifier        9 / 122\n",
      "\tSVC                           3 / 111\n",
      "\tXGBClassifier                 5 / 129\n",
      "\n",
      "LogisticRegression                    W / L\n",
      "--------------------\n",
      "\tAdaBoostClassifier            58 / 58\n",
      "\tBernoulliNB                   93 / 24\n",
      "\tDecisionTreeClassifier        42 / 85\n",
      "\tExtraTreesClassifier          10 / 116\n",
      "\tGaussianNB                    132 / 14\n",
      "\tGradientBoostingClassifier    7 / 122\n",
      "\tKNeighborsClassifier          55 / 70\n",
      "\tLinearSVC                     25 / 6\n",
      "\tMultinomialNB                 139 / 1\n",
      "\tPassiveAggressiveClassifier   82 / 15\n",
      "\tRandomForestClassifier        10 / 118\n",
      "\tSVC                           6 / 102\n",
      "\tXGBClassifier                 6 / 128\n",
      "\n",
      "MultinomialNB                         W / L\n",
      "--------------------\n",
      "\tAdaBoostClassifier            22 / 126\n",
      "\tBernoulliNB                   25 / 107\n",
      "\tDecisionTreeClassifier        10 / 137\n",
      "\tExtraTreesClassifier          2 / 157\n",
      "\tGaussianNB                    66 / 77\n",
      "\tGradientBoostingClassifier    2 / 156\n",
      "\tKNeighborsClassifier          10 / 137\n",
      "\tLinearSVC                     3 / 137\n",
      "\tLogisticRegression            1 / 139\n",
      "\tPassiveAggressiveClassifier   8 / 124\n",
      "\tRandomForestClassifier        2 / 154\n",
      "\tSVC                           1 / 151\n",
      "\tXGBClassifier                 2 / 157\n",
      "\n",
      "PassiveAggressiveClassifier           W / L\n",
      "--------------------\n",
      "\tAdaBoostClassifier            51 / 73\n",
      "\tBernoulliNB                   85 / 38\n",
      "\tDecisionTreeClassifier        35 / 99\n",
      "\tExtraTreesClassifier          5 / 133\n",
      "\tGaussianNB                    129 / 18\n",
      "\tGradientBoostingClassifier    3 / 137\n",
      "\tKNeighborsClassifier          40 / 90\n",
      "\tLinearSVC                     19 / 68\n",
      "\tLogisticRegression            15 / 82\n",
      "\tMultinomialNB                 124 / 8\n",
      "\tRandomForestClassifier        9 / 133\n",
      "\tSVC                           1 / 128\n",
      "\tXGBClassifier                 1 / 139\n",
      "\n",
      "RandomForestClassifier                W / L\n",
      "--------------------\n",
      "\tAdaBoostClassifier            118 / 9\n",
      "\tBernoulliNB                   143 / 1\n",
      "\tDecisionTreeClassifier        105 / 0\n",
      "\tExtraTreesClassifier          28 / 27\n",
      "\tGaussianNB                    158 / 1\n",
      "\tGradientBoostingClassifier    20 / 37\n",
      "\tKNeighborsClassifier          120 / 4\n",
      "\tLinearSVC                     122 / 9\n",
      "\tLogisticRegression            118 / 10\n",
      "\tMultinomialNB                 154 / 2\n",
      "\tPassiveAggressiveClassifier   133 / 9\n",
      "\tSVC                           59 / 34\n",
      "\tXGBClassifier                 16 / 48\n",
      "\n",
      "SVC                                   W / L\n",
      "--------------------\n",
      "\tAdaBoostClassifier            101 / 15\n",
      "\tBernoulliNB                   138 / 5\n",
      "\tDecisionTreeClassifier        83 / 26\n",
      "\tExtraTreesClassifier          27 / 59\n",
      "\tGaussianNB                    154 / 5\n",
      "\tGradientBoostingClassifier    24 / 67\n",
      "\tKNeighborsClassifier          111 / 8\n",
      "\tLinearSVC                     111 / 3\n",
      "\tLogisticRegression            102 / 6\n",
      "\tMultinomialNB                 151 / 1\n",
      "\tPassiveAggressiveClassifier   128 / 1\n",
      "\tRandomForestClassifier        34 / 59\n",
      "\tXGBClassifier                 21 / 74\n",
      "\n",
      "XGBClassifier                         W / L\n",
      "--------------------\n",
      "\tAdaBoostClassifier            132 / 0\n",
      "\tBernoulliNB                   151 / 0\n",
      "\tDecisionTreeClassifier        124 / 1\n",
      "\tExtraTreesClassifier          51 / 22\n",
      "\tGaussianNB                    158 / 1\n",
      "\tGradientBoostingClassifier    34 / 12\n",
      "\tKNeighborsClassifier          130 / 4\n",
      "\tLinearSVC                     129 / 5\n",
      "\tLogisticRegression            128 / 6\n",
      "\tMultinomialNB                 157 / 2\n",
      "\tPassiveAggressiveClassifier   139 / 1\n",
      "\tRandomForestClassifier        48 / 16\n",
      "\tSVC                           74 / 21\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from itertools import product\n",
    "\n",
    "for model1 in all_models:\n",
    "    print('{}{}W / L'.format(model1,\n",
    "                               ' ' * (38 - len(model1))))\n",
    "    print('--------------------')\n",
    "    for model2 in all_models:\n",
    "        if model1 == model2:\n",
    "            continue\n",
    "        print('\\t{}{}{} / {}'.format(model2,\n",
    "                                     ' ' * (30 - len(model2)),\n",
    "                                     model_tourneys[(model1, model2)],\n",
    "                                     model_tourneys[(model2, model1)]))\n",
    "    print('')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from itertools import product\n",
    "import numpy as np\n",
    "\n",
    "model_tourney_matrix = []\n",
    "\n",
    "for pair in list(product(all_models, all_models)):\n",
    "    model_tourney_matrix.append(model_tourneys[pair])\n",
    "    \n",
    "model_tourney_matrix = np.array(model_tourney_matrix).reshape((14, 14))\n",
    "all_models = list(np.array(all_models)[np.argsort(model_tourney_matrix.sum(axis=1))[::-1]])\n",
    "model_tourney_matrix = model_tourney_matrix[:, np.argsort(model_tourney_matrix.sum(axis=1))[::-1]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from itertools import product\n",
    "import numpy as np\n",
    "\n",
    "\"\"\"\n",
    "TODO: This code snippet needs to be run twice (this is the 2nd time it's being run here,\n",
    "with the above snippet being the first time) so the chart below will display properly.\n",
    "There's some weird bug going on here that we should dig in to.\n",
    "\"\"\"\n",
    "\n",
    "model_tourney_matrix = []\n",
    "\n",
    "for pair in list(product(all_models, all_models)):\n",
    "    model_tourney_matrix.append(model_tourneys[pair])\n",
    "    \n",
    "model_tourney_matrix = np.array(model_tourney_matrix).reshape((14, 14))\n",
    "all_models = list(np.array(all_models)[np.argsort(model_tourney_matrix.sum(axis=1))[::-1]])\n",
    "model_tourney_matrix = model_tourney_matrix[:, np.argsort(model_tourney_matrix.sum(axis=1))[::-1]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "''"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAqQAAAHbCAYAAAAd2oggAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngNQ555xzzpXU/wOTv7PX/hlOhAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12fa24438>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sb\n",
    "model_nice_dict = {\n",
    "    'AdaBoostClassifier': 'AB',\n",
    "    'BernoulliNB': 'BNB',\n",
    "    'LogisticRegression': 'LR',\n",
    "    'MultinomialNB': 'MNB',\n",
    "    'PassiveAggressiveClassifier': 'PA',\n",
    "    'SGDClassifier': 'SGD',\n",
    "    'GaussianNB': 'GNB',\n",
    "    'DecisionTreeClassifier': 'DT',\n",
    "    'ExtraTreesClassifier': 'ERF',\n",
    "    'RandomForestClassifier': 'RF',\n",
    "    'GradientBoostingClassifier':'GTB',\n",
    "    'KNeighborsClassifier': 'KNN',\n",
    "    'SVC': 'SVM',\n",
    "    'XGBClassifier': 'XGB',\n",
    "    'LinearSVC': 'LSVM'\n",
    "}\n",
    "model_nice_dict_y = {\n",
    "    'AdaBoostClassifier': 'AdaBoost',\n",
    "    'BernoulliNB': 'Bernoulli Naive Bayes',\n",
    "    'LogisticRegression': 'Logistic Regression',\n",
    "    'MultinomialNB': 'Multinomial Naive Bayes',\n",
    "    'PassiveAggressiveClassifier': 'Passive Aggressive',\n",
    "    'SGDClassifier': 'Linear Model trained via\\nStochastic Gradient Descent',\n",
    "    'GaussianNB': 'Gaussian Naive Bayes',\n",
    "    'DecisionTreeClassifier': 'Decision Tree',\n",
    "    'ExtraTreesClassifier': 'Extra Random Forest',\n",
    "    'RandomForestClassifier': 'Random Forest',\n",
    "    'GradientBoostingClassifier':'Gradient Tree Boosting',\n",
    "    'KNeighborsClassifier': 'K-Nearest Neighbors',\n",
    "    'SVC': 'Support Vector Machine',\n",
    "    'XGBClassifier': 'XGBoost',\n",
    "    'LinearSVC': 'Linear Support Vector Machine'\n",
    "}\n",
    "\n",
    "model_nice = []\n",
    "model_nice_y = []\n",
    "for m in all_models:\n",
    "    model_nice.append(model_nice_dict[m])\n",
    "    model_nice_y.append(model_nice_dict_y[m])\n",
    "    \n",
    "plt.figure(figsize=(10, 10))\n",
    "sb.heatmap(model_tourney_matrix, square=True, annot=True, fmt='0.0f', cmap='Purples',\n",
    "           xticklabels=model_nice, yticklabels=model_nice_y, cbar_kws={'shrink': 0.56})\n",
    "plt.xticks(fontsize=12)\n",
    "plt.yticks(fontsize=12)\n",
    "plt.xlabel('Losses', fontsize=14)\n",
    "plt.ylabel('Wins', fontsize=14)\n",
    "plt.title('How many times model X outperformed model Y (out of 165)', fontsize=18)\n",
    "h = plt.gcf()\n",
    "plt.tight_layout()\n",
    "#h.savefig('figs/model_outperformance.pdf', bbox_inches='tight')\n",
    "#plt.savefig('figures/sklearn-model-x-outperform-model-y.pdf', bbox_inches='tight')\n",
    ";"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# How many models do we need to cover all data sets?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "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>dataset</th>\n",
       "      <th>classifier</th>\n",
       "      <th>accuracy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...</td>\n",
       "      <td>AdaBoostClassifier</td>\n",
       "      <td>0.504</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...</td>\n",
       "      <td>BernoulliNB</td>\n",
       "      <td>0.512</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...</td>\n",
       "      <td>DecisionTreeClassifier</td>\n",
       "      <td>0.527</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...</td>\n",
       "      <td>ExtraTreesClassifier</td>\n",
       "      <td>0.536</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...</td>\n",
       "      <td>GaussianNB</td>\n",
       "      <td>0.501</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                             dataset              classifier  \\\n",
       "0  GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...      AdaBoostClassifier   \n",
       "1  GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...             BernoulliNB   \n",
       "2  GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...  DecisionTreeClassifier   \n",
       "3  GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...    ExtraTreesClassifier   \n",
       "4  GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...              GaussianNB   \n",
       "\n",
       "   accuracy  \n",
       "0     0.504  \n",
       "1     0.512  \n",
       "2     0.527  \n",
       "3     0.536  \n",
       "4     0.501  "
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import pdb\n",
    "\n",
    "data = pd.read_csv('sklearn-benchmark6-data.tsv.gz', sep='\\t', names=['dataset',\n",
    "                                                                     'classifier',\n",
    "                                                                     'parameters',\n",
    "                                                                     'accuracy', \n",
    "                                                                     'macrof1',\n",
    "                                                                     'bal_accuracy']).fillna('')\n",
    "\n",
    "data = data.groupby(['dataset','classifier'])['accuracy'].max().reset_index()\n",
    "data['accuracy'] = data['accuracy'].apply(lambda x: round(x, 3))\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/randal_olson/anaconda/lib/python3.6/site-packages/ipykernel_launcher.py:9: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  if __name__ == '__main__':\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model & Data Set Coverage \\\\ \\hline\n",
      "XGBClassifier & 123 \\\\\n",
      "GradientBoostingClassifier & 111 \\\\\n",
      "ExtraTreesClassifier & 99 \\\\\n",
      "RandomForestClassifier & 91 \\\\\n",
      "SVC & 81 \\\\\n",
      "DecisionTreeClassifier & 35 \\\\\n",
      "LogisticRegression & 29 \\\\\n",
      "LinearSVC & 28 \\\\\n",
      "AdaBoostClassifier & 24 \\\\\n",
      "KNeighborsClassifier & 20 \\\\\n",
      "PassiveAggressiveClassifier & 15 \\\\\n",
      "BernoulliNB & 8 \\\\\n",
      "GaussianNB & 4 \\\\\n",
      "MultinomialNB & 2 \\\\\n"
     ]
    }
   ],
   "source": [
    "from collections import defaultdict\n",
    "from tqdm import tqdm\n",
    "import numpy as np\n",
    "\n",
    "dataset_best_models = defaultdict(list)\n",
    "model_counts = defaultdict(int)\n",
    "\n",
    "for dataset, group_dataset in data.groupby('dataset'):\n",
    "    group_dataset['accuracy'] /= group_dataset['accuracy'].max()\n",
    "    dataset_best_models[dataset] = group_dataset.loc[\n",
    "        group_dataset['accuracy'] >= 0.99, 'classifier'].values\n",
    "\n",
    "for dataset in dataset_best_models:\n",
    "    for model in dataset_best_models[dataset]:\n",
    "        model_counts[model] += 1\n",
    "print('Model','&','Data Set Coverage','\\\\\\\\ \\\\hline')\n",
    "for model in sorted(model_counts, key=model_counts.get, reverse=True):\n",
    "    print(model,'&',model_counts[model],'\\\\\\\\')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/randal_olson/anaconda/lib/python3.6/site-packages/ipykernel_launcher.py:5: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  \"\"\"\n",
      "/Users/randal_olson/anaconda/lib/python3.6/site-packages/ipykernel_launcher.py:30: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n"
     ]
    }
   ],
   "source": [
    "dataset_best_models = defaultdict(list)\n",
    "model_counts = defaultdict(int)\n",
    "\n",
    "for dataset, group_dataset in data.groupby('dataset'):\n",
    "    group_dataset['accuracy'] /= group_dataset['accuracy'].max()\n",
    "    dataset_best_models[dataset] = group_dataset.loc[\n",
    "        group_dataset['accuracy'] >= 0.99, 'classifier'].values\n",
    "\n",
    "for dataset in dataset_best_models:\n",
    "    for model in dataset_best_models[dataset]:\n",
    "        model_counts[model] += 1\n",
    "\n",
    "dataset_exclude_set = set()\n",
    "top_models = []\n",
    "\n",
    "while len(dataset_exclude_set) != len(data['dataset'].unique()):\n",
    "    next_top_model = sorted(model_counts, key=model_counts.get, reverse=True)[0]\n",
    "    top_models.append((model_counts[next_top_model], next_top_model))\n",
    "    \n",
    "    for dataset in dataset_best_models:\n",
    "        if next_top_model in dataset_best_models[dataset]:\n",
    "            dataset_exclude_set.add(dataset)\n",
    "\n",
    "    dataset_best_models = defaultdict(list)\n",
    "    model_counts = defaultdict(int)\n",
    "    \n",
    "    for dataset, group_dataset in data.groupby('dataset'):\n",
    "        if dataset in dataset_exclude_set:\n",
    "            continue\n",
    "        group_dataset['accuracy'] /= group_dataset['accuracy'].max()\n",
    "        dataset_best_models[dataset] = group_dataset.loc[\n",
    "            group_dataset['accuracy'] >= 0.99, 'classifier'].values\n",
    "        \n",
    "    for dataset in dataset_best_models:\n",
    "        for model in dataset_best_models[dataset]:\n",
    "            model_counts[model] += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(123, 'XGBClassifier'),\n",
       " (22, 'ExtraTreesClassifier'),\n",
       " (9, 'SVC'),\n",
       " (7, 'GradientBoostingClassifier'),\n",
       " (1, 'LinearSVC'),\n",
       " (1, 'RandomForestClassifier'),\n",
       " (1, 'KNeighborsClassifier')]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "top_models"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# How many model-parameter combinations do we need to cover all data sets?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "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>dataset</th>\n",
       "      <th>classifier</th>\n",
       "      <th>parameters</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>classifier-params</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...</td>\n",
       "      <td>AdaBoostClassifier</td>\n",
       "      <td>learning_rate=0.01,n_estimators=10,random_stat...</td>\n",
       "      <td>0.481</td>\n",
       "      <td>AdaBoostClassifier-learning_rate=0.01,n_estima...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...</td>\n",
       "      <td>AdaBoostClassifier</td>\n",
       "      <td>learning_rate=0.01,n_estimators=100,random_sta...</td>\n",
       "      <td>0.485</td>\n",
       "      <td>AdaBoostClassifier-learning_rate=0.01,n_estima...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...</td>\n",
       "      <td>AdaBoostClassifier</td>\n",
       "      <td>learning_rate=0.01,n_estimators=50,random_stat...</td>\n",
       "      <td>0.482</td>\n",
       "      <td>AdaBoostClassifier-learning_rate=0.01,n_estima...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...</td>\n",
       "      <td>AdaBoostClassifier</td>\n",
       "      <td>learning_rate=0.01,n_estimators=500,random_sta...</td>\n",
       "      <td>0.486</td>\n",
       "      <td>AdaBoostClassifier-learning_rate=0.01,n_estima...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...</td>\n",
       "      <td>AdaBoostClassifier</td>\n",
       "      <td>learning_rate=0.1,n_estimators=10,random_state...</td>\n",
       "      <td>0.477</td>\n",
       "      <td>AdaBoostClassifier-learning_rate=0.1,n_estimat...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                             dataset          classifier  \\\n",
       "0  GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...  AdaBoostClassifier   \n",
       "1  GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...  AdaBoostClassifier   \n",
       "2  GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...  AdaBoostClassifier   \n",
       "3  GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...  AdaBoostClassifier   \n",
       "4  GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...  AdaBoostClassifier   \n",
       "\n",
       "                                          parameters  accuracy  \\\n",
       "0  learning_rate=0.01,n_estimators=10,random_stat...     0.481   \n",
       "1  learning_rate=0.01,n_estimators=100,random_sta...     0.485   \n",
       "2  learning_rate=0.01,n_estimators=50,random_stat...     0.482   \n",
       "3  learning_rate=0.01,n_estimators=500,random_sta...     0.486   \n",
       "4  learning_rate=0.1,n_estimators=10,random_state...     0.477   \n",
       "\n",
       "                                   classifier-params  \n",
       "0  AdaBoostClassifier-learning_rate=0.01,n_estima...  \n",
       "1  AdaBoostClassifier-learning_rate=0.01,n_estima...  \n",
       "2  AdaBoostClassifier-learning_rate=0.01,n_estima...  \n",
       "3  AdaBoostClassifier-learning_rate=0.01,n_estima...  \n",
       "4  AdaBoostClassifier-learning_rate=0.1,n_estimat...  "
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import pdb\n",
    "\n",
    "data = pd.read_csv('sklearn-benchmark6-data.tsv.gz', sep='\\t', names=['dataset',\n",
    "                                                                     'classifier',\n",
    "                                                                     'parameters',\n",
    "                                                                     'accuracy', \n",
    "                                                                     'macrof1',\n",
    "                                                                     'bal_accuracy']).fillna('')\n",
    "\n",
    "data = data.groupby(['dataset','classifier','parameters'])['accuracy'].max().reset_index()\n",
    "\n",
    "data['classifier-params'] = data['classifier'].values + '-' + data['parameters'].values\n",
    "data['accuracy'] = data['accuracy'].apply(lambda x: round(x, 3))\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/randal_olson/anaconda/lib/python3.6/site-packages/pandas/core/indexing.py:517: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  self.obj[item] = s\n"
     ]
    }
   ],
   "source": [
    "from collections import defaultdict \n",
    "\n",
    "dataset_best_models = defaultdict(list)\n",
    "model_counts = defaultdict(int)\n",
    "\n",
    "for dataset, group_dataset in data.groupby('dataset'):\n",
    "    group_dataset.loc[:, 'accuracy'] = group_dataset['accuracy'].values / group_dataset['accuracy'].max()\n",
    "    dataset_best_models[dataset] = group_dataset.loc[\n",
    "        group_dataset['accuracy'] >= 0.99, 'classifier-params'].values\n",
    "\n",
    "for dataset in dataset_best_models:\n",
    "    for model in dataset_best_models[dataset]:\n",
    "        model_counts[model] += 1\n",
    "\n",
    "dataset_exclude_set = set()\n",
    "top_models = []\n",
    "\n",
    "while len(dataset_exclude_set) != len(data['dataset'].unique()):\n",
    "    next_top_model = sorted(model_counts, key=model_counts.get, reverse=True)[0]\n",
    "    top_models.append((model_counts[next_top_model], next_top_model))\n",
    "    \n",
    "    if len(top_models) == 10:\n",
    "        break\n",
    "    \n",
    "    # Don't allow repeat models\n",
    "    data = data.loc[data['classifier'] != next_top_model.split('-')[0].strip()]\n",
    "    \n",
    "    for dataset in dataset_best_models:\n",
    "        if next_top_model in dataset_best_models[dataset]:\n",
    "            dataset_exclude_set.add(dataset)\n",
    "\n",
    "    dataset_best_models = defaultdict(list)\n",
    "    model_counts = defaultdict(int)\n",
    "    \n",
    "    for dataset, group_dataset in data.groupby('dataset'):\n",
    "        if dataset in dataset_exclude_set:\n",
    "            continue\n",
    "        group_dataset.loc[:, 'accuracy'] = group_dataset.loc[:, 'accuracy'].values / group_dataset['accuracy'].max()\n",
    "        dataset_best_models[dataset] = group_dataset.loc[\n",
    "            group_dataset['accuracy'] >= 0.99, 'classifier-params'].values\n",
    "        \n",
    "    for dataset in dataset_best_models:\n",
    "        for model in dataset_best_models[dataset]:\n",
    "            model_counts[model] += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(55,\n",
       "  'GradientBoostingClassifier-n_estimators=500,min_impurity_decrease=0.0,max_features=sqrt,learning_rate=0.1,loss=deviance,random_state=324089'),\n",
       " (13,\n",
       "  'ExtraTreesClassifier-n_estimators=500,min_impurity_decrease=0.0,max_features=0.25,criterion=entropy,random_state=324089'),\n",
       " (8,\n",
       "  'XGBClassifier-n_estimators=500,learning_rate=0.1,gamma=0.30000000000000004,max_depth=2,subsample=0.9,seed=324089,nthread=1'),\n",
       " (8,\n",
       "  'LogisticRegression-C=1.5,penalty=l1,fit_intercept=True,dual=False,random_state=324089'),\n",
       " (7,\n",
       "  'RandomForestClassifier-n_estimators=100,min_impurity_decrease=0.0,max_features=0.25,criterion=gini,random_state=324089'),\n",
       " (8,\n",
       "  'SVC-C=0.1,gamma=0.5,kernel=poly,degree=2,coef0=50.0,random_state=324089'),\n",
       " (8,\n",
       "  'LinearSVC-C=0.2,penalty=l1,fit_intercept=True,dual=False,random_state=324089'),\n",
       " (7, 'KNeighborsClassifier-n_neighbors=10,weights=distance'),\n",
       " (6,\n",
       "  'AdaBoostClassifier-learning_rate=1.0,n_estimators=500,random_state=324089'),\n",
       " (8,\n",
       "  'PassiveAggressiveClassifier-C=0.01,loss=hinge,fit_intercept=True,random_state=324089')]"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "top_models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('PassiveAggressiveClassifier-C=0.01,loss=hinge,fit_intercept=True,random_state=324089',\n",
       "  8),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.0075,max_features=None,criterion=entropy,random_state=324089',\n",
       "  6),\n",
       " ('PassiveAggressiveClassifier-C=0.01,loss=squared_hinge,fit_intercept=True,random_state=324089',\n",
       "  5),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.0,max_features=None,criterion=entropy,random_state=324089',\n",
       "  5),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.01,max_features=None,criterion=entropy,random_state=324089',\n",
       "  5),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.003,max_features=0.5,criterion=entropy,random_state=324089',\n",
       "  4),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.00375,max_features=0.5,criterion=entropy,random_state=324089',\n",
       "  4),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.0035,max_features=0.75,criterion=gini,random_state=324089',\n",
       "  4),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.00475,max_features=None,criterion=entropy,random_state=324089',\n",
       "  4),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.005,max_features=None,criterion=entropy,random_state=324089',\n",
       "  4),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.015,max_features=None,criterion=entropy,random_state=324089',\n",
       "  4),\n",
       " ('PassiveAggressiveClassifier-C=0.001,loss=squared_hinge,fit_intercept=True,random_state=324089',\n",
       "  4),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.00125,max_features=0.5,criterion=gini,random_state=324089',\n",
       "  3),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.00025,max_features=None,criterion=entropy,random_state=324089',\n",
       "  3),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.0005,max_features=None,criterion=entropy,random_state=324089',\n",
       "  3),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.00075,max_features=None,criterion=entropy,random_state=324089',\n",
       "  3),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.001,max_features=None,criterion=entropy,random_state=324089',\n",
       "  3),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.00125,max_features=None,criterion=entropy,random_state=324089',\n",
       "  3),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.0015,max_features=None,criterion=entropy,random_state=324089',\n",
       "  3),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.00175,max_features=None,criterion=entropy,random_state=324089',\n",
       "  3),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.002,max_features=0.75,criterion=gini,random_state=324089',\n",
       "  3),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.002,max_features=None,criterion=entropy,random_state=324089',\n",
       "  3),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.0022500000000000003,max_features=0.75,criterion=gini,random_state=324089',\n",
       "  3),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.0022500000000000003,max_features=None,criterion=entropy,random_state=324089',\n",
       "  3),\n",
       " ('DecisionTreeClassifier-min_impurity_decrease=0.0025,max_features=None,criterion=entropy,random_state=324089',\n",
       "  3)]"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "[(x, model_counts[x]) for x in sorted(model_counts, key=model_counts.get, reverse=True)[:25]]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model/data biclustering analysis\n",
    "\n",
    "Create matrix of data sets vs. best model accuracy on those data sets.\n",
    "\n",
    "Cluster the matrix.\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "\n",
    "\n",
    "data = pd.read_csv('sklearn-benchmark6-data.tsv.gz', sep='\\t', names=['dataset',\n",
    "                                                                     'classifier',\n",
    "                                                                     'parameters',\n",
    "                                                                     'accuracy', \n",
    "                                                                     'macrof1',\n",
    "                                                                     'bal_accuracy']).fillna('')\n",
    "\n",
    "data = data.groupby(['classifier', 'dataset'])['bal_accuracy'].max().reset_index()\n",
    "# print(\"classifiers before drop:\",data['classifier'].unique())\n",
    "# data = data[data['classifier']!='LinearSVC']\n",
    "# data = data[data['classifier']!='SVC']\n",
    "print('loaded ',data['dataset'].unique().shape[0],'datasets and ', data['classifier'].unique().shape[0],'classifiers')\n",
    "# data['classifier-params'] = data['classifier'].values + '-' + data['parameters'].values\n",
    "data['bal_accuracy'] = data['bal_accuracy'].apply(lambda x: round(x, 3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from tqdm import tqdm\n",
    "from sklearn.cluster.bicluster import SpectralBiclustering\n",
    "from sklearn.metrics import consensus_score\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "import pdb\n",
    "# get model results with best parameters \n",
    "\n",
    "all_models = np.asarray((data['classifier'].unique()))\n",
    "all_datasets = np.asarray((data['dataset'].unique()))\n",
    "model_data_acc = np.zeros([len(all_models),len(all_datasets)])\n",
    "model_data_acc_norm = np.zeros([len(all_models),len(all_datasets)])\n",
    "\n",
    "ranks = np.zeros([len(all_models),len(all_datasets)])\n",
    "#print(\"model_data_acc.shape:\",model_data_acc.shape)\n",
    "# all_models = []\n",
    "for i,(clf, group_clf) in enumerate(tqdm(data.groupby('dataset'))):\n",
    "    print('clf:',clf)\n",
    "    data_acc  = group_clf['bal_accuracy'].transpose()\n",
    "    model_data_acc[:,i] = data_acc\n",
    "#     model_data_acc_norm[:,i] = (data_acc-data_acc.min())/(data_acc.max() - data_acc.min())\n",
    "    model_data_acc_norm[:,i] = (data_acc-data_acc.mean())\n",
    "\n",
    "#     all_models.append(clf)\n",
    "    \n",
    "all_models = np.asarray(all_models)\n",
    "# get ranks\n",
    "for i,mda in enumerate(model_data_acc.transpose()):\n",
    "    #print(\"mda shape:\",mda.shape)\n",
    "    temp = mda.argsort()\n",
    "    ranks[temp,i] = np.arange(len(mda))\n",
    "np.savetxt('figs/ranks',ranks)\n",
    "np.savetxt('figs/model_data_accuracy',model_data_acc)\n",
    "\n",
    "\n",
    "print(\"clustering...\")\n",
    "np.random.seed(42)\n",
    "model = SpectralBiclustering(n_clusters=(4,10), random_state=0)\n",
    "model.fit(model_data_acc)\n",
    "#score = consensus_score(model.biclusters_, (rows[:, row_idx], columns[:, col_idx]))\n",
    "\n",
    "#print(\"consensus score: {:.3f}\".format(score))\n",
    "r = model.row_labels_[:]\n",
    "c = model.column_labels_[:]\n",
    "#rearrange data into biclusters\n",
    "fit_data_norm = model_data_acc_norm[np.argsort(r)]\n",
    "fit_data_norm = fit_data_norm[:, np.argsort(c)]\n",
    "\n",
    "fit_data = model_data_acc[np.argsort(r)]\n",
    "fit_data = fit_data[:, np.argsort(c)]\n",
    "\n",
    "all_models = all_models[np.argsort(r)]\n",
    "all_datasets = all_datasets[np.argsort(c)]\n",
    "print('done')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "model_nice_dict = {\n",
    "    'AdaBoostClassifier': 'AdaBoost',\n",
    "    'BernoulliNB': 'Bernoulli NB',\n",
    "    'LinearSVC': 'Linear SVC',\n",
    "    'LogisticRegression': 'Logistic Regression',\n",
    "    'MultinomialNB': 'Multinomial NB',\n",
    "    'PassiveAggressiveClassifier': 'Passive Aggressive',\n",
    "    'SGDClassifier': 'SGD',\n",
    "    'GaussianNB': 'Gaussian NB',\n",
    "    'DecisionTreeClassifier': 'Decision Tree',\n",
    "    'ExtraTreesClassifier': 'Extra Trees',\n",
    "    'RandomForestClassifier': 'Random Forest',\n",
    "    'GradientBoostingClassifier':'Gradient Boosting',\n",
    "    'KNeighborsClassifier': 'K-Nearest Neighbor',\n",
    "    'SVC': 'SVC',\n",
    "    'XGBClassifier': 'XGBoost',\n",
    "    'LinearSVC': 'Linear Support Vector Machine'\n",
    "}\n",
    "model_nice = []\n",
    "for m in all_models:\n",
    "    model_nice.append(model_nice_dict[m])\n",
    "    \n",
    "    \n",
    "print(\"biclusters_:\",len(model.biclusters_))\n",
    "#plot\n",
    "# h = plt.figure(figsize=(4,3),sharey=True)\n",
    "# ax = plt.subplot(111)\n",
    "h,ax = plt.subplots(3,figsize=(10,9))\n",
    "# ax = h.add_subplot(311)\n",
    "tmp = ax[0].imshow(fit_data[:,:],cmap=plt.cm.RdBu) \n",
    "# ax[0].set_title('A')\n",
    "# ax[0].set_xlabel('A')\n",
    "cbar=plt.colorbar(tmp,ax=ax[0],orientation='vertical',shrink=0.8)\n",
    "cbar.set_label('Balanced Accuracy')\n",
    "ax[0].set_yticks(range(len(all_models))) #,rotation=90\n",
    "ax[0].set_yticklabels(model_nice) #,rotation=90\n",
    "# ax[1].set_xlabel('Data Set',size=16)\n",
    "ax[0].set_xticks(np.arange(len(all_datasets),step=10))\n",
    "ax[0].xaxis.tick_top()\n",
    "\n",
    "# h = plt.gcf()\n",
    "# ax = plt.gca( )\n",
    "ax[0].set_aspect(4)\n",
    "\n",
    "# h.tight_layout()\n",
    "# h = plt.gcf()\n",
    "\n",
    "# h.savefig(\"figs/bicluster.pdf\",dpi=100)\n",
    "\n",
    "# k = plt.figure(figsize=(10,3))\n",
    "# ax = h.add_subplot(312)\n",
    "tmp = ax[1].matshow(fit_data_norm[:,:],cmap=plt.cm.RdBu) \n",
    "\n",
    "cbar=plt.colorbar(tmp,ax=ax[1],orientation='vertical',shrink=0.8)\n",
    "cbar.set_label('Deviation from Mean')\n",
    "ax[1].set_yticks(range(len(all_models))) #,rotation=90\n",
    "ax[1].set_yticklabels(model_nice) #,rotation=90\n",
    "# ax[1].set_xlabel('Data Set',size=16)\n",
    "ax[1].set_xticks(np.arange(len(all_datasets),step=10))\n",
    "# ax[1].set_xlabel('B')\n",
    "# h = plt.gcf()\n",
    "# ax = plt.gca( )\n",
    "ax[1].set_aspect(4)\n",
    "# h.tight_layout()\n",
    "# h = plt.gcf()\n",
    "\n",
    "# k.savefig(\"figs/bicluster_zeromean.pdf\",dpi=100)\n",
    "\n",
    "\n",
    "# h2 = plt.figure(figsize=(10,3))\n",
    "# ax = h.add_subplot(313)\n",
    "cluster_labels  = np.outer(np.sort(model.row_labels_) + 1,np.sort(model.column_labels_) + 1)\n",
    "boundary = np.zeros((cluster_labels.shape[0],cluster_labels.shape[1]))\n",
    "for i,cr in enumerate(cluster_labels[1:]):\n",
    "    for j,cc in enumerate(cr[1:]):\n",
    "        if cluster_labels[i-1,j] != cluster_labels[i,j]:\n",
    "            boundary[i,j] = 1\n",
    "        if cluster_labels[i,j-1] != cluster_labels[i,j]:\n",
    "            boundary[i,j] = 1\n",
    "            \n",
    "tmp=ax[2].matshow(cluster_labels,cmap=plt.cm.Purples,alpha=1)\n",
    "# tmp = \n",
    "# ydata = [0,165,0,165,0,165]\n",
    "# tmp=ax[2].plot((0,165),(2.5,2.5))\n",
    "# plt.gca().invert_yaxis()\n",
    "cbar=plt.colorbar(tmp,ax=ax[2],orientation='vertical',shrink=0.8)\n",
    "cbar.set_label('Bicluster ID')\n",
    "plt.yticks(range(len(all_models)), model_nice) #,rotation=90\n",
    "ax[2].set_xlabel('Dataset',size=16)\n",
    "plt.xticks(np.arange(len(all_datasets),step=10))\n",
    "# ax[2].set_xlabel('C')\n",
    "# h = plt.gcf()\n",
    "# ax = plt.gca( )\n",
    "ax[2].set_aspect(4)\n",
    "h.tight_layout()\n",
    "# plt.subplots_adjust(top=0.95)\n",
    "#h.savefig(\"figs/cluster_all.pdf\",dpi=100)\n",
    "#h.savefig(\"figs/cluster_all.eps\",dpi=100)\n",
    "#h.savefig(\"figs/cluster_all.png\",dpi=100)\n",
    "\n",
    "\n",
    "plt.show()\n",
    "j=0\n",
    "for c in np.unique(cluster_labels[0,:]):\n",
    "    print('cluster',c,':')\n",
    "    for d in all_datasets[cluster_labels[0,:]==c]:\n",
    "#         print('',j,\":\",d)\n",
    "        print('\"'+d+'\"',end=',')\n",
    "        j+=1\n",
    "    print(' ')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# How do the algorithms cluster?\n",
    "\n",
    "Create matrix of data sets vs. median model accuracy on those data sets.\n",
    "\n",
    "Cluster the matrix using Agglomerative Clustering. Look at the resulting dendrogram."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "\n",
    "data = pd.read_csv('sklearn-benchmark6-data.tsv.gz', sep='\\t', names=['dataset',\n",
    "                                                                     'classifier',\n",
    "                                                                     'parameters',\n",
    "                                                                     'accuracy', \n",
    "                                                                     'macrof1',\n",
    "                                                                     'bal_accuracy']).fillna('')\n",
    "\n",
    "# data = data.groupby(['classifier', 'dataset', 'parameters'])['accuracy'].mean().reset_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from tqdm import tqdm\n",
    "from scipy.cluster import hierarchy\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "all_models = np.asarray(sorted(data['classifier'].unique()))\n",
    "# remove SVC from all_models\n",
    "\n",
    "all_datasets = np.asarray(sorted(data['dataset'].unique()))\n",
    "\n",
    "model_data_acc = np.zeros([len(all_models),len(all_datasets)])\n",
    "ranks = np.zeros([len(all_models),len(all_datasets)])\n",
    "#print(\"model_data_acc.shape:\",model_data_acc.shape)\n",
    "all_models = []\n",
    "\n",
    "for i,(clf, group_clf) in enumerate(tqdm(data.groupby('classifier'))):\n",
    "#     if clf != 'SVC':\n",
    "    model_best_params_acc = np.zeros(len(all_datasets))\n",
    "    # find best parameter setings for model, based on median cv score for each parameter setting\n",
    "    for params,group_clf_params in group_clf.groupby('parameters'):\n",
    "        # across data sets\n",
    "        for j,a in enumerate(group_clf_params.groupby('dataset')['accuracy'].median()):\n",
    "            if a > model_best_params_acc[j]:\n",
    "                model_best_params_acc[j] = a\n",
    "    # model i's accuracy is the median cv accuracy of the best parameter settings for that model, across data sets\n",
    "    model_data_acc[i,:] = model_best_params_acc\n",
    "\n",
    "    all_models.append(clf)\n",
    "\n",
    "all_models = np.asarray(all_models)\n",
    "\n",
    "# get ranks\n",
    "for i,mda in enumerate(model_data_acc.transpose()):\n",
    "    #print(\"mda shape:\",mda.shape)\n",
    "    temp = mda.argsort()\n",
    "    ranks[temp,i] = np.arange(len(mda))\n",
    "    \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model_nice_dict = {\n",
    "    'AdaBoostClassifier': 'AdaBoost',\n",
    "    'BernoulliNB': 'Bernoulli NB',\n",
    "    'LinearSVC': 'Linear SVC',\n",
    "    'LogisticRegression': 'Logistic Regression',\n",
    "    'MultinomialNB': 'Multinomial NB',\n",
    "    'PassiveAggressiveClassifier': 'Passive Aggressive',\n",
    "    'SGDClassifier': 'SGD',\n",
    "    'GaussianNB': 'Gaussian NB',\n",
    "    'DecisionTreeClassifier': 'Decision Tree',\n",
    "    'ExtraTreesClassifier': 'Extra Trees',\n",
    "    'RandomForestClassifier': 'Random Forest',\n",
    "    'GradientBoostingClassifier':'Gradient Boosting',\n",
    "    'KNeighborsClassifier': 'K-Nearest Neighbor',\n",
    "    'SVC': 'SVC',\n",
    "    'XGBClassifier': 'XGBoost',\n",
    "    'LinearSVC': 'Linear Support Vector Machine'\n",
    "}\n",
    "model_nice = []\n",
    "for m in all_models:\n",
    "    model_nice.append(model_nice_dict[m])\n",
    "    \n",
    "print(\"clustering...\")\n",
    "Z = hierarchy.linkage(ranks, 'single')\n",
    "dn = hierarchy.dendrogram(Z, labels = model_nice, orientation='right')\n",
    "plt.title('Ranking Dendrogram')\n",
    "h = plt.gcf()\n",
    "plt.figure(1,figsize=(6,4))\n",
    "plt.tight_layout()\n",
    "#h.savefig('figs/HAC_models_ranking.pdf')\n",
    "\n",
    "plt.figure()\n",
    "Z = hierarchy.linkage(model_data_acc, 'single')\n",
    "dn = hierarchy.dendrogram(Z, labels = model_nice, orientation='right')\n",
    "plt.title('Accuracy Dendrogram')\n",
    "h = plt.gcf()\n",
    "plt.figure(1,figsize=(6,4))\n",
    "plt.tight_layout()\n",
    "#h.savefig('figs/HAC_models_accuracy.pdf')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# How do the data sets cluster?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.cluster import KMeans\n",
    "from sklearn.metrics import silhouette_score, silhouette_samples\n",
    "from sklearn.decomposition import PCA\n",
    "import numpy\n",
    "from tqdm import tqdm\n",
    "#==========\n",
    "# optimal K via elbow method with silhouette score which produces a better elbow.\n",
    "#==========\n",
    "X = model_data_acc.transpose()\n",
    "Ks = np.arange(2,147,1)\n",
    "Inertias = []\n",
    "Silhouettes = []\n",
    "np.random.seed(2)\n",
    "# loop through k values\n",
    "for K in tqdm(Ks):\n",
    "    km = KMeans(n_clusters=K, init='k-means++',copy_x=False).fit(X)\n",
    "    labels = km.labels_\n",
    "    centers = km.cluster_centers_\n",
    "    inertia = km.inertia_\n",
    "    Silhouettes.append(silhouette_score(X,labels))\n",
    "    # Inertias[K-1] = km.inertia_\n",
    "    Inertias.append(km.inertia_)\n",
    "\n",
    "# line plot of K versus Silhouette score with best value marked with x \n",
    "plt.figure(1)\n",
    "plt.plot(Ks,Silhouettes,label='silhouette')\n",
    "plt.plot(Ks[np.argmax(Silhouettes)],Silhouettes[np.argmax(Silhouettes)],marker = 'o',color='r',markersize=7)\n",
    "plt.text(Ks[np.argmax(Silhouettes)]-2,Silhouettes[np.argmax(Silhouettes)],\"K = \"+repr(Ks[np.argmax(Silhouettes)]))\n",
    "plt.ylim(0.95*np.min(Silhouettes),1.05*np.max(Silhouettes))\n",
    "plt.ylabel(\"Average silhouette score\") #Y-axis label\n",
    "plt.xlabel(\"K\") #X-axis label\n",
    "plt.title(\"Choice of K\") #Plot title\n",
    "plt.tight_layout()\n",
    "plt.savefig(\"figs/k_silhouette.pdf\")\n",
    "\n",
    "plt.figure(2)\n",
    "plt.plot(Ks,Inertias,label='inertia')\n",
    "plt.plot(Ks[np.argmin(Inertias)],Inertias[np.argmin(Inertias)],marker = 'o',color='r',markersize=7)\n",
    "plt.text(Ks[np.argmin(Inertias)]-2,Inertias[np.argmin(Inertias)],\"K = \"+repr(Ks[np.argmin(Inertias)]))\n",
    "plt.ylim(0.95*np.min(Inertias),1.05*np.max(Inertias))\n",
    "plt.ylabel(\"Inertias\") #Y-axis label\n",
    "plt.xlabel(\"K\") #X-axis label\n",
    "plt.title(\"Choice of K\") #Plot title\n",
    "plt.tight_layout()\n",
    "plt.savefig(\"figs/k_inertia.pdf\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# =====\n",
    "# plot cluster centers on 2 principal component axes\n",
    "# =====\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.cluster import KMeans\n",
    "import itertools\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "\n",
    "marker =('+', 'x', 'o', '*','s','^','<','v','>') \n",
    "h = plt.figure()\n",
    "ss = StandardScaler()\n",
    "X = ss.fit_transform(model_data_acc.transpose())\n",
    "\n",
    "pca = PCA(n_components = 2)\n",
    "X_pca = pca.fit_transform(X)\n",
    "nc=4\n",
    "unique_classes = np.array((0,1,2,3)) \n",
    "km = KMeans(n_clusters=nc, init='k-means++',copy_x=False,random_state=0).fit(X)\n",
    "labels = km.labels_\n",
    "centers = km.cluster_centers_\n",
    "unique_labels = np.unique(labels)\n",
    "# centers_pca = centers\n",
    "centers_pca = pca.transform(centers)\n",
    "colors = plt.cm.Dark2(np.linspace(0, 1, len(unique_labels)))\n",
    "\n",
    "for k, col in zip(unique_labels, colors):\n",
    "    label_mask = (k==labels)\n",
    "    xy = X_pca[label_mask]\n",
    "    plt.plot(xy[:,0], xy[:, 1], linestyle = '', marker=marker[k%nc], markerfacecolor=col, markersize=5, alpha=1)\n",
    "\n",
    "for k, col in zip(unique_labels, colors): \n",
    "     plt.plot(centers_pca[k,0],centers_pca[k,1], linestyle='', marker=marker[k%nc], markerfacecolor=col,markersize=20,alpha=0.3)\n",
    "\n",
    "plt.xlabel('PC 1')\n",
    "plt.ylabel('PC 2')\n",
    "plt.tight_layout()\n",
    "h.savefig('figs/k_means_PCA_data.pdf')\n",
    "\n",
    "h2 = plt.figure()\n",
    "features = model_nice\n",
    "    \n",
    "for k,col in zip(unique_labels,colors):\n",
    "    label_mask = (k==labels)\n",
    "    coverage = np.sum(label_mask)\n",
    "    xk_mean = np.mean(ss.inverse_transform(X[label_mask]),axis=0)\n",
    "    offset = k*0.1-np.mean(np.unique(labels))*0.1\n",
    "    print('col:',col)\n",
    "    plt.bar(np.arange(len(features))+offset,xk_mean,align='center',width=0.1,facecolor=col,label='cluster '+marker[k%nc]+' ('+str(coverage)+' instances)')\n",
    "\n",
    "plt.ylim(0,1.1)\n",
    "plt.gca().set_xticks(np.arange(len(features)))\n",
    "plt.gca().set_xticklabels(list(features),fontsize=8,rotation=90)\n",
    "plt.legend(loc=3,fontsize=6) #(bbox_to_anchor=(1.05, 1), \n",
    "plt.tight_layout()\n",
    "h2.savefig('figs/data_ml_bar.pdf')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Comparison of tuned to un-tuned results\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from tqdm import tqdm\n",
    "import pandas as pd\n",
    "\n",
    "data = pd.read_csv('sklearn-benchmark6-data.tsv.gz', sep='\\t', names=['dataset',\n",
    "                                                                     'classifier',\n",
    "                                                                     'parameters',\n",
    "                                                                     'accuracy', \n",
    "                                                                     'macrof1',\n",
    "                                                                     'bal_accuracy']).fillna('')\n",
    "\n",
    "data = data.groupby(['classifier', 'dataset', 'parameters'])['accuracy'].mean().reset_index()\n",
    "data['accuracy'] = data['accuracy'].apply(lambda x: round(x, 3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "clf_defaults_dict = {\n",
    "    'GradientBoostingClassifier': 'n_estimators=100,min_impurity_decrease=0.0,max_features=None,learning_rate=0.1,loss=deviance,random_state=324089',\n",
    "    'RandomForestClassifier': 'n_estimators=10,min_impurity_decrease=0.0,max_features=sqrt,criterion=gini,random_state=324089',\n",
    "    'SVC': 'C=1.0,gamma=auto,kernel=rbf,degree=2,coef0=0.0,random_state=324089',\n",
    "    'ExtraTreesClassifier': 'n_estimators=10,min_impurity_decrease=0.0,max_features=sqrt,criterion=gini,random_state=324089',\n",
    "    'KNeighborsClassifier': 'n_neighbors=5,weights=uniform',\n",
    "    'LogisticRegression': 'C=1.0,penalty=l2,fit_intercept=True,dual=False,random_state=324089',\n",
    "    'DecisionTreeClassifier': 'min_impurity_decrease=0.0,max_features=None,criterion=gini,random_state=324089',\n",
    "    'PassiveAggressiveClassifier': 'C=1.0,loss=hinge,fit_intercept=False,random_state=324089',\n",
    "    'AdaBoostClassifier': 'learning_rate=1.0,n_estimators=50,random_state=324089',\n",
    "    'BernoulliNB': 'alpha=1.0,fit_prior=True,binarize=0.0',\n",
    "    'GaussianNB': '',\n",
    "    'MultinomialNB': 'alpha=1.0,fit_prior=True',\n",
    "    'XGBClassifier': 'n_estimators=100,learning_rate=0.1,gamma=0.0,max_depth=3,subsample=1.0,seed=324089,nthread=1',\n",
    "    'LinearSVC': 'C=1.0,penalty=l2,fit_intercept=True,dual=True,random_state=324089'\n",
    "}\n",
    "\n",
    "default_params_list = ['-'.join([k, v]) for k, v in clf_defaults_dict.items()]\n",
    "\n",
    "default_scores = data.loc[\n",
    "    data.apply(\n",
    "        lambda record: '-'.join([record['classifier'],\n",
    "                                 record['parameters']]) in default_params_list,\n",
    "        axis=1)].drop('parameters', axis=1)\n",
    "\n",
    "best_scores = data.groupby(['dataset', 'classifier'])['accuracy'].max().reset_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "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>classifier</th>\n",
       "      <th>dataset</th>\n",
       "      <th>accuracy_default</th>\n",
       "      <th>accuracy_best</th>\n",
       "      <th>accuracy_default_scaled</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>AdaBoost</td>\n",
       "      <td>GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...</td>\n",
       "      <td>0.477</td>\n",
       "      <td>0.504</td>\n",
       "      <td>0.027</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>AdaBoost</td>\n",
       "      <td>GAMETES_Epistasis_2-Way_20atts_0.1H_EDM-1_1</td>\n",
       "      <td>0.496</td>\n",
       "      <td>0.515</td>\n",
       "      <td>0.019</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>AdaBoost</td>\n",
       "      <td>GAMETES_Epistasis_2-Way_20atts_0.4H_EDM-1_1</td>\n",
       "      <td>0.474</td>\n",
       "      <td>0.500</td>\n",
       "      <td>0.026</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>AdaBoost</td>\n",
       "      <td>GAMETES_Epistasis_3-Way_20atts_0.2H_EDM-1_1</td>\n",
       "      <td>0.526</td>\n",
       "      <td>0.532</td>\n",
       "      <td>0.006</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>AdaBoost</td>\n",
       "      <td>GAMETES_Heterogeneity_20atts_1600_Het_0.4_0.2_...</td>\n",
       "      <td>0.477</td>\n",
       "      <td>0.510</td>\n",
       "      <td>0.033</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  classifier                                            dataset  \\\n",
       "0   AdaBoost  GAMETES_Epistasis_2-Way_1000atts_0.4H_EDM-1_ED...   \n",
       "1   AdaBoost        GAMETES_Epistasis_2-Way_20atts_0.1H_EDM-1_1   \n",
       "2   AdaBoost        GAMETES_Epistasis_2-Way_20atts_0.4H_EDM-1_1   \n",
       "3   AdaBoost        GAMETES_Epistasis_3-Way_20atts_0.2H_EDM-1_1   \n",
       "4   AdaBoost  GAMETES_Heterogeneity_20atts_1600_Het_0.4_0.2_...   \n",
       "\n",
       "   accuracy_default  accuracy_best  accuracy_default_scaled  \n",
       "0             0.477          0.504                    0.027  \n",
       "1             0.496          0.515                    0.019  \n",
       "2             0.474          0.500                    0.026  \n",
       "3             0.526          0.532                    0.006  \n",
       "4             0.477          0.510                    0.033  "
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "merged_scores = default_scores.merge(best_scores,\n",
    "                                     on=['classifier', 'dataset'],\n",
    "                                     suffixes=['_default', '_best'])\n",
    "\n",
    "merged_scores['accuracy_default_scaled'] = merged_scores['accuracy_best'] - merged_scores['accuracy_default']\n",
    "\n",
    "model_names_dict = {\n",
    "    'AdaBoostClassifier': 'AdaBoost',\n",
    "    'BernoulliNB': 'Bernoulli Naive Bayes',\n",
    "    'LogisticRegression': 'Logistic Regression',\n",
    "    'MultinomialNB': 'Multinomial Naive Bayes',\n",
    "    'PassiveAggressiveClassifier': 'Passive Aggressive',\n",
    "    'GaussianNB': 'Gaussian Naive Bayes',\n",
    "    'DecisionTreeClassifier': 'Decision Tree',\n",
    "    'ExtraTreesClassifier': 'Extra Random Forest',\n",
    "    'RandomForestClassifier': 'Random Forest',\n",
    "    'GradientBoostingClassifier':'Gradient Tree Boosting',\n",
    "    'KNeighborsClassifier': 'K-Nearest Neighbors',\n",
    "    'SVC': 'Support Vector Machine',\n",
    "    'XGBClassifier': 'XGBoost',\n",
    "    'LinearSVC': 'Linear Support Vector Machine'\n",
    "}\n",
    "\n",
    "merged_scores['classifier'] = merged_scores['classifier'].apply(lambda x: model_names_dict[x])\n",
    "\n",
    "merged_scores.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "''"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAtIAAAK/CAYAAABEACouAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1464e5198>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import seaborn as sb\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "sb.set_style('whitegrid')\n",
    "\n",
    "plt.figure(figsize=(9, 12))\n",
    "sb.boxplot(data=merged_scores.sort_values('classifier'),\n",
    "           y='classifier', x='accuracy_default_scaled', notch=True,\n",
    "           palette=[sb.color_palette('Purples', n_colors=2)[1]])\n",
    "plt.ylabel('')\n",
    "plt.xlabel('10-fold CV accuracy improvement by tuning models', fontsize=16)\n",
    "plt.xticks(fontsize=14)\n",
    "plt.yticks(fontsize=14)\n",
    "plt.xlim(0., 0.5)\n",
    "#plt.title('Tuning machine learning model parameters almost always improves\\nmodel performance', fontsize=22)\n",
    "#plt.savefig('figs/tuned_untuned_accuracy_boxplot.pdf', bbox_inches='tight')\n",
    ";"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# print model abbreviation table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model_nice_dict = {\n",
    "    'AdaBoostClassifier': 'AB',\n",
    "    'BernoulliNB': 'BNB',\n",
    "    'LinearSVC': 'LSVC',\n",
    "    'LogisticRegression': 'LR',\n",
    "    'MultinomialNB': 'MNB',\n",
    "    'PassiveAggressiveClassifier': 'PAC',\n",
    "    'SGDClassifier': 'SGD',\n",
    "    'GaussianNB': 'GNB',\n",
    "    'DecisionTreeClassifier': 'DT',\n",
    "    'ExtraTreesClassifier': 'ET',\n",
    "    'RandomForestClassifier': 'RF',\n",
    "    'GradientBoostingClassifier':'GB',\n",
    "    'KNeighborsClassifier': 'KNN',\n",
    "    'XGBClassifier': 'XGBoost',\n",
    "    'LinearSVC': 'Linear Support Vector Machine'\n",
    "}\n",
    "model_nice = []\n",
    "for m in model_nice_dict:\n",
    "    print(model_nice_dict)\n",
    "\n",
    "\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "\n",
    "data = pd.read_csv('sklearn-benchmark6-data.tsv.gz', sep='\\t', names=['dataset',\n",
    "                                                                     'classifier',\n",
    "                                                                     'parameters',\n",
    "                                                                     'accuracy', \n",
    "                                                                     'macrof1',\n",
    "                                                                     'bal_accuracy']).fillna('')\n",
    "data = data.groupby(['dataset','classifier'])['bal_accuracy'].max().reset_index()\n",
    "\n",
    "print(data[:5])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import seaborn as sns\n",
    "\n",
    "datan = data.groupby(['dataset']).max().sort_values('bal_accuracy',ascending=False)\n",
    "# print(datan)\n",
    "\n",
    "datamv = data.groupby(['dataset']).median()+data.groupby(['dataset']).var()\n",
    "datamv = datamv.sort_values('bal_accuracy',ascending=True)\n",
    "\n",
    "# for i in datamv.index:\n",
    "#     print(data[i,'bal_accuracy'])\n",
    "print(datamv[::2])\n",
    "# print(datan.index)\n",
    "\n",
    "print(data['dataset'][:5])\n",
    "plt.figure(figsize=(10,5))\n",
    "sns.set_style(\"whitegrid\")\n",
    "s=sns.boxplot(data=data,x='dataset',y='bal_accuracy',order=datan.index,fliersize=3,linewidth=0.75)\n",
    "s.set_xticks(np.arange(len(np.unique(data['dataset'])),step=10))\n",
    "s.set_xticklabels(np.arange(len(np.unique(data['dataset'])),step=10))\n",
    "yticks = np.hstack((np.arange(0.6,step=0.1),np.arange(0.6,1.05,step=0.05)))\n",
    "s.set_yticks(yticks)\n",
    "s.set_yticklabels(['{0:.2f}'.format(x) for x in yticks],size=9)\n",
    "plt.ylim(0,1.1)\n",
    "plt.ylabel('Balanced Accuracy',size=16)\n",
    "plt.xlabel('Dataset',size=16)\n",
    "h = plt.gcf()\n",
    "h.tight_layout()\n",
    "#h.savefig('figs/boxplot_all.pdf',bbox_inches='tight')\n",
    "#h.savefig('figs/boxplot_all.png',bbox_inches='tight')\n",
    "\n",
    "\n",
    "print('90% cutoff:',len(datan[datan['bal_accuracy']>=0.9]))\n",
    "print('80% cutoff:',len(datan[datan['bal_accuracy']>=0.8]))\n",
    "print('70% cutoff:',len(datan[datan['bal_accuracy']>=0.7]))\n",
    "print('60% cutoff:',len(datan[datan['bal_accuracy']>=0.6]))\n",
    "\n",
    "# for i,d in enumerate(datan.index):\n",
    "\n",
    "#     print('data set ',i,':',data['dataset'][d])\n",
    "# plt.gca().set_xticks(np.arange(len(data),step=10))\n",
    "# plt.gca().set_xticklabels(str(np.arange(len(data),step=10)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "cutoff = np.empty(100)\n",
    "for i in np.arange(100):\n",
    "    cutoff[i] = len(datan[datan['bal_accuracy']>=i/100])/len(datan)\n",
    "plt.bar(np.arange(len(cutoff)),cutoff)\n",
    "plt.xlim(50,100)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "149./165"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "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": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}