Lisbon Data Science Meetup - Hyperparameter Search.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Machine Learning Hyperparameter Search (from Grid Search to SKOPT)\n",
    "This is the Jupyter (IPython) notebook support for the Lisbon Data Science meetup of 2016/10/19, held at Startup Lisboa, Portugal. Link to the Meetup [here](http://www.meetup.com/Lisbon-Open-Data-Meetup/events/234758716/?_af=event&_af_eid=234758716&https=off).\n",
    "\n",
    "### Introduction\n",
    "An important part of using any machine learning algorithm (such as Logistic Regression or Random Forests) is tuning the hyper parameters. \n",
    "\n",
    "Tools such as scikit's **Grid-Search** and **Randomized Search** have greatly improved the ability to use CPU instead of sweat and tears to tune over large hyperparameter spaces. \n",
    "\n",
    "More recently, a new library has appeared - **Scikit-Optimize** - which allows us to use Bayesian Optimization (among other things) to help find good hyperparameters with less computation resources.  \n",
    "\n",
    "---\n",
    "**Contents**:\n",
    "- Grid-Search CV\n",
    "- Randomized Search CV\n",
    "- Scikit-Optimize (SKOPT)\n",
    "\n",
    "---\n",
    "**Pre-requisites**:\n",
    "- Some Python programming language background\n",
    "- Machine Learning (supervised learning) models such as Random Forests\n",
    "- Some exposure to Pandas, Numpy and Scikit-learn\n",
    "- Motivation to have the best running classifier in the room\n",
    "\n",
    "---\n",
    "**Required Python lybraries** (latest versions if possible, you can easily install them with ***pip***)\n",
    "- sklearn\n",
    "- skopt\n",
    "- pandas\n",
    "- numpy\n",
    "- scipy\n",
    "- matplotlib\n",
    "\n",
    "---  \n",
    "Presenter: Hugo Lopes ([LinkedIn](https://linkedin.com/in/hugodlopes))  \n",
    "Contact: hjdlopes@gmail.com"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Initializations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "plt.rcParams.update({'font.size': 16})\n",
    "\n",
    "import pandas as pd \n",
    "import numpy as np\n",
    "from scipy.stats import randint\n",
    "\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.model_selection import GridSearchCV, RandomizedSearchCV\n",
    "from sklearn.model_selection import StratifiedKFold\n",
    "from sklearn.model_selection import cross_val_score\n",
    "from sklearn.metrics import roc_auc_score\n",
    "\n",
    "from skopt import gp_minimize"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "np.random.seed(seed=1)\n",
    "results = {}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load Data\n",
    "Go to [Give Me Some Credit](https://www.kaggle.com/c/GiveMeSomeCredit/data) (you need to be registered at Kaggle to download the data, if you're not, now it is the best time to do it :) ). Download the [*cs-training*](https://www.kaggle.com/c/GiveMeSomeCredit/download/cs-training.csv) (.csv (7.21 mb)) and place it in the same folder of this notebook."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "data = pd.read_csv('cs-training.csv', index_col=0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's check the structure of the data, now without NaN."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>SeriousDlqin2yrs</th>\n",
       "      <th>RevolvingUtilizationOfUnsecuredLines</th>\n",
       "      <th>age</th>\n",
       "      <th>NumberOfTime30-59DaysPastDueNotWorse</th>\n",
       "      <th>DebtRatio</th>\n",
       "      <th>MonthlyIncome</th>\n",
       "      <th>NumberOfOpenCreditLinesAndLoans</th>\n",
       "      <th>NumberOfTimes90DaysLate</th>\n",
       "      <th>NumberRealEstateLoansOrLines</th>\n",
       "      <th>NumberOfTime60-89DaysPastDueNotWorse</th>\n",
       "      <th>NumberOfDependents</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>0.766127</td>\n",
       "      <td>45</td>\n",
       "      <td>2</td>\n",
       "      <td>0.802982</td>\n",
       "      <td>9120.0</td>\n",
       "      <td>13</td>\n",
       "      <td>0</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "      <td>2.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>0.957151</td>\n",
       "      <td>40</td>\n",
       "      <td>0</td>\n",
       "      <td>0.121876</td>\n",
       "      <td>2600.0</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>0.658180</td>\n",
       "      <td>38</td>\n",
       "      <td>1</td>\n",
       "      <td>0.085113</td>\n",
       "      <td>3042.0</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>0.233810</td>\n",
       "      <td>30</td>\n",
       "      <td>0</td>\n",
       "      <td>0.036050</td>\n",
       "      <td>3300.0</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0</td>\n",
       "      <td>0.907239</td>\n",
       "      <td>49</td>\n",
       "      <td>1</td>\n",
       "      <td>0.024926</td>\n",
       "      <td>63588.0</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0</td>\n",
       "      <td>0.213179</td>\n",
       "      <td>74</td>\n",
       "      <td>0</td>\n",
       "      <td>0.375607</td>\n",
       "      <td>3500.0</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0</td>\n",
       "      <td>0.305682</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>5710.000000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>8</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0</td>\n",
       "      <td>0.754464</td>\n",
       "      <td>39</td>\n",
       "      <td>0</td>\n",
       "      <td>0.209940</td>\n",
       "      <td>3500.0</td>\n",
       "      <td>8</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   SeriousDlqin2yrs  RevolvingUtilizationOfUnsecuredLines  age  \\\n",
       "1                 1                              0.766127   45   \n",
       "2                 0                              0.957151   40   \n",
       "3                 0                              0.658180   38   \n",
       "4                 0                              0.233810   30   \n",
       "5                 0                              0.907239   49   \n",
       "6                 0                              0.213179   74   \n",
       "7                 0                              0.305682   57   \n",
       "8                 0                              0.754464   39   \n",
       "\n",
       "   NumberOfTime30-59DaysPastDueNotWorse    DebtRatio  MonthlyIncome  \\\n",
       "1                                     2     0.802982         9120.0   \n",
       "2                                     0     0.121876         2600.0   \n",
       "3                                     1     0.085113         3042.0   \n",
       "4                                     0     0.036050         3300.0   \n",
       "5                                     1     0.024926        63588.0   \n",
       "6                                     0     0.375607         3500.0   \n",
       "7                                     0  5710.000000            NaN   \n",
       "8                                     0     0.209940         3500.0   \n",
       "\n",
       "   NumberOfOpenCreditLinesAndLoans  NumberOfTimes90DaysLate  \\\n",
       "1                               13                        0   \n",
       "2                                4                        0   \n",
       "3                                2                        1   \n",
       "4                                5                        0   \n",
       "5                                7                        0   \n",
       "6                                3                        0   \n",
       "7                                8                        0   \n",
       "8                                8                        0   \n",
       "\n",
       "   NumberRealEstateLoansOrLines  NumberOfTime60-89DaysPastDueNotWorse  \\\n",
       "1                             6                                     0   \n",
       "2                             0                                     0   \n",
       "3                             0                                     0   \n",
       "4                             0                                     0   \n",
       "5                             1                                     0   \n",
       "6                             1                                     0   \n",
       "7                             3                                     0   \n",
       "8                             0                                     0   \n",
       "\n",
       "   NumberOfDependents  \n",
       "1                 2.0  \n",
       "2                 1.0  \n",
       "3                 0.0  \n",
       "4                 0.0  \n",
       "5                 0.0  \n",
       "6                 1.0  \n",
       "7                 0.0  \n",
       "8                 0.0  "
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head(8)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's perform a fast and simple way to eliminate missing data by just dropping the rows with NaN. There are other methods to avoid wasting these rows, but it is out of the scope of this notebook."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of entries without NaN:\n",
      "---------------------------\n",
      "SeriousDlqin2yrs                        150000\n",
      "RevolvingUtilizationOfUnsecuredLines    150000\n",
      "age                                     150000\n",
      "NumberOfTime30-59DaysPastDueNotWorse    150000\n",
      "DebtRatio                               150000\n",
      "MonthlyIncome                           120269\n",
      "NumberOfOpenCreditLinesAndLoans         150000\n",
      "NumberOfTimes90DaysLate                 150000\n",
      "NumberRealEstateLoansOrLines            150000\n",
      "NumberOfTime60-89DaysPastDueNotWorse    150000\n",
      "NumberOfDependents                      146076\n",
      "dtype: int64\n"
     ]
    }
   ],
   "source": [
    "print('Number of entries without NaN:')\n",
    "print('---------------------------')\n",
    "print(data.count())\n",
    "data.dropna(inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now it is time to select the **target** (*y*, the first column) and **features** (*X*, the remaining columns). Note, we will perform binary classification so target is either 0 or 1.\n",
    "\n",
    "Let us also **reduce the size of the dataset to 10000 rows (clients)** so that we can perform computations live."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of features (columns): 10\n",
      "Number of loans (rows): 10000\n"
     ]
    }
   ],
   "source": [
    "# Get target column\n",
    "y = data['SeriousDlqin2yrs']\n",
    "\n",
    "# Get features columns\n",
    "X = data.ix[:, 1:]\n",
    "\n",
    "# Shuffle data and reduce dataset\n",
    "idx = np.random.permutation(y.size)\n",
    "X = X.iloc[idx]\n",
    "X = X.iloc[:10000]\n",
    "y = y.iloc[idx]\n",
    "y = y.iloc[:10000]\n",
    "\n",
    "print('Number of features (columns):', X.shape[1])\n",
    "print('Number of loans (rows):', X.shape[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train-Test Split\n",
    "Before we construct our model, let's divide the data into a **training dataset** (80 percent of the data) and a **test dataset** (20 percent of the data). We will train the classifier on the training dataset, and then evaluate the performance of our classifier with the test dataset. \n",
    "\n",
    "This is very useful to check if we are **overfitting** our training data, i.e., getting very good results within the training dataset and bad results with our testing dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.20, random_state=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Making a First Prediction (with classifier's default configuration)\n",
    "Let's make use of the [**Random Forest Classifier**](http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestClassifier.html) (available in [scikit-learn](http://scikit-learn.org/stable/)).  \n",
    "\n",
    "First, let's train it with the default configuration, then evaluate its performance on the test data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
       "            max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
       "            min_impurity_split=1e-07, min_samples_leaf=1,\n",
       "            min_samples_split=2, min_weight_fraction_leaf=0.0,\n",
       "            n_estimators=10, n_jobs=-1, oob_score=False, random_state=1,\n",
       "            verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf = RandomForestClassifier(n_jobs=-1, random_state=1)\n",
    "clf.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From the output above we can check the configuration of the default Random Forests, such as the number of trees in the forest, *n_estimators=10*. So, with the created and trained classifier, let's check the **ROC AUC** on the given test data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test ROC AUC: 0.7907\n"
     ]
    }
   ],
   "source": [
    "#results['Default'] = clf.score(X_test, y_test)\n",
    "results['Default'] = roc_auc_score(y_test, clf.predict_proba(X_test)[:, 1])\n",
    "print('Test ROC AUC: %.4f' % results['Default'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Tuning our Classifier: Grid-Search Cross-Validation\n",
    "As scikit-learn defines it, [**GridSearchCV**](http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.GridSearchCV.html#sklearn.model_selection.GridSearchCV) is: *Exhaustive search over specified parameter values for an estimator.*  \n",
    "The user defines a set of values for the classifier hyperparameters. Then, the GridSearchCV implements a “fit” and a “score” method. The evaluation is performed with Cross-Validation."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 1. Create Cross-Validation (CV) Folds\n",
    "This [**StratifiedKFold**](http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.StratifiedKFold.html) cross-validation object is a variation of KFold that returns stratified folds, ie., the folds are made by preserving the percentage of samples for each class (stratification is generally a better scheme, both in terms of bias and variance)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "skfold_cv = StratifiedKFold(n_splits=3, shuffle=True, random_state=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2. Fixed Parameters\n",
    "Secondly, let us define some parameters that we won't vary in our Random Forest ([click here](http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestClassifier.html) for full definition of its hyperparameters):\n",
    "- n_jobs=-1, to use all cores\n",
    "- random_state=1, so that we can re-run experiements with same results later\n",
    "- n_estimators=50, the number of trees in the forest, usually the higher the better (so does not make much sense to optimize this...). 50 is not a high number for this experiment, but the higher the value is, the more computational resources we'll need."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Fixed parameters\n",
    "params_fixed = {\n",
    "    'n_jobs': -1,\n",
    "    'random_state': 1,\n",
    "    'n_estimators': 100\n",
    "}\n",
    "# Initialize Random Forest model with the fixed parameters\n",
    "clf = RandomForestClassifier(**params_fixed)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3. Scoring metric\n",
    "Now, we have to decide our **scoring metric** to evaluate the performance of our classifier. Some examples are *accuracy*, *mean_squared_error*, *roc_auc*."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#score_metric = 'accuracy'\n",
    "score_metric = 'roc_auc'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### The Grid-Search CV\n",
    "The Random Forest Classifier has a bunch of hyperparameters we can tune ([click here](http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestClassifier.html) for full description), thus for sake of learning how to work with the optimization tools we'll not go into their full description. The tuning process is analogous to any scikit-learn model, but with different hyperparameters.\n",
    "\n",
    "For example, let's pick the following ones:\n",
    "- max_depth\n",
    "- max_features\n",
    "- min_samples_split\n",
    "- min_samples_leaf\n",
    "- criterion"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 32 candidates, totalling 96 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done  42 tasks      | elapsed:   12.3s\n",
      "[Parallel(n_jobs=-1)]: Done  96 out of  96 | elapsed:   32.1s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Done!\n"
     ]
    }
   ],
   "source": [
    "# Define the grid for the selected hyperparameters\n",
    "params_grid = {\"max_depth\": [1, 20],\n",
    "              \"max_features\": [1, 10],\n",
    "              \"min_samples_split\": [1, 10],\n",
    "              \"min_samples_leaf\": [1, 10],\n",
    "              \"criterion\": [\"gini\", \"entropy\"]}\n",
    "\n",
    "# Run Grid-Search CV\n",
    "results_gs = GridSearchCV(estimator=clf,\n",
    "                          param_grid=params_grid,\n",
    "                          cv=skfold_cv,\n",
    "                          scoring=score_metric,\n",
    "                          n_jobs=-1,\n",
    "                          verbose=1)\n",
    "results_gs.fit(X, y)\n",
    "print('Done!')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let us evaluate the results of all runs:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA1cAAAGYCAYAAACu4xi4AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nZmZmZtYGTq7MzMzMzMza4P8DQnIS+4sG/JAAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x108dac630>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(14,6))\n",
    "plt.plot(results_gs.cv_results_['mean_test_score'], 'o-')\n",
    "plt.grid()\n",
    "plt.xlabel('Iteration number (in GridSearchCV)')\n",
    "plt.ylabel('ROC AUC')\n",
    "plt.title('GridSearchCV - Results')\n",
    "plt.xlim([0, len(results_gs.cv_results_['mean_test_score'])-1])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "What are the parameters that led to the best accuracy that was achieved?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'criterion': 'gini',\n",
       " 'max_depth': 20,\n",
       " 'max_features': 1,\n",
       " 'min_samples_leaf': 10,\n",
       " 'min_samples_split': 1}"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results_gs.best_params_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "What is the best ROC AUC we got?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test ROC AUC: 0.8549\n",
      "(Test ROC AUC: 0.7907 - Default configuration)\n"
     ]
    }
   ],
   "source": [
    "results['GridSearchCV'] = results_gs.best_score_\n",
    "print('Test ROC AUC: %.4f' % results['GridSearchCV'])\n",
    "print('(Test ROC AUC: %.4f - Default configuration)' % results['Default'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "There is improvement here, which means that the default configuration is not the best one (as expected...)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Tuning our Classifier: Randomized Search CV\n",
    "The [**RandomizedSearchCV**](http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.RandomizedSearchCV.html#sklearn.model_selection.RandomizedSearchCV) implements a randomized search over parameters, where each setting is sampled from a distribution over possible parameter values (i.e., not all parameter values are tried out).  \n",
    "\n",
    "Besides the distribution of the parameters (***param_distributions***) we must set the number of interations (***n_iter***) we would like the search to run, which trades-off runtime vs quality of the solution. The remaining parameters are very similar to GridSearchCV.\n",
    "\n",
    "RandomizedSearchCV has **two main benefits** over GridSearchCV:\n",
    "- A budget (number of iterations) can be chosen independent of the number of parameters and possible values.\n",
    "- Adding parameters that do not influence the performance does not decrease efficiency."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 20 candidates, totalling 60 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done  42 tasks      | elapsed:   17.9s\n",
      "[Parallel(n_jobs=-1)]: Done  60 out of  60 | elapsed:   23.6s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Done!\n"
     ]
    }
   ],
   "source": [
    "# Distribution of the parameters (randint: discrete random variables)\n",
    "param_dist = {\"max_depth\": randint(1, 21),\n",
    "              \"max_features\": randint(1, 11),\n",
    "              \"min_samples_split\": randint(1, 11),\n",
    "              \"min_samples_leaf\": randint(1, 11),\n",
    "              \"criterion\": [\"gini\", \"entropy\"]}\n",
    "\n",
    "# Number of iterations to perform\n",
    "n_iter = 20\n",
    "\n",
    "# Run Randomized Search\n",
    "results_rgs = RandomizedSearchCV(estimator=clf,\n",
    "                                param_distributions=param_dist,\n",
    "                                n_iter=n_iter,\n",
    "                                cv=skfold_cv,           # use same CV folds as in GridSearchCV\n",
    "                                scoring=score_metric,   # use same scoring metric as in GridSearchCV\n",
    "                                random_state=1,\n",
    "                                n_jobs=-1,\n",
    "                                verbose=1)\n",
    "results_rgs.fit(X, y)\n",
    "print('Done!')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let us evaluate the results of all runs:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA2EAAAGYCAYAAADLDurHAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1046832b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(14,6))\n",
    "plt.plot(results_rgs.cv_results_['mean_test_score'], 'ro-')\n",
    "plt.grid()\n",
    "plt.xlabel('Iteration number')\n",
    "plt.ylabel('ROC AUC')\n",
    "plt.title('RandomizedSearchCV - Results')\n",
    "plt.xlim([0, len(results_rgs.cv_results_['mean_test_score'])-1])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "What are the parameters that led to the best accuracy that was achieved?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'criterion': 'gini',\n",
       " 'max_depth': 10,\n",
       " 'max_features': 2,\n",
       " 'min_samples_leaf': 8,\n",
       " 'min_samples_split': 1}"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results_rgs.best_params_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "What is the best accuracy we got?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test ROC AUC: 0.8551\n",
      "(Test ROC AUC of Grid-Search: 0.8549)\n"
     ]
    }
   ],
   "source": [
    "results['RandomizedGridSearchCV'] = results_rgs.best_score_\n",
    "print('Test ROC AUC: %.4f' % results['RandomizedGridSearchCV'])\n",
    "print('(Test ROC AUC of Grid-Search: %.4f)' % results['GridSearchCV'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "There is a very slight improvement here, but was it just mere luck?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Tuning our Classifier: Scikit-Optimize (SKOPT)\n",
    "[**SKOPT**](https://scikit-optimize.github.io/) is a sequential model-based optimization. SKOPT hopefully results in equivalent or better solutions, but within less evaluations (less time consuming). It can also be reusable in various contexts besides scikit-learn (e.g., minimizing a function).\n",
    "\n",
    "The library is built on top of NumPy, SciPy and Scikit-Learn.  \n",
    "\n",
    "First step is to define the **objective function** we want to minimize (it accepts an array of parameters and return the function values). In our case, it is the **cross-validation negative (since we are minimizing!) ROC AUC over our dataset**, as a function of its hyperparameters. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def objective_func(params):\n",
    "    # Get the parameters\n",
    "    max_depth, max_features, min_samples_split, min_samples_leaf, criterion = params\n",
    "    # Set the parameters in the model\n",
    "    clf.set_params(max_depth=max_depth, \n",
    "                     max_features=max_features,\n",
    "                     min_samples_split=min_samples_split,\n",
    "                     min_samples_leaf=min_samples_leaf,\n",
    "                     criterion=criterion)\n",
    "    # Perform CV and return the mean of the scoring values\n",
    "    skfold_cv_splits = skfold_cv.split(X, y)\n",
    "    return -np.mean(cross_val_score(clf,\n",
    "                                    X,\n",
    "                                    y, \n",
    "                                    cv=skfold_cv_splits, \n",
    "                                    n_jobs=-1, \n",
    "                                    scoring=score_metric))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Next step is to define the **bounds of the parameters** in the search space we want to explore (we could also set the starting point) and the **number of iterations**."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Optimization Space (parameters and their boundaries):\n",
    "space = [(1, 20),               # max_depth\n",
    "         (1, 10),               # max_features \n",
    "         (1, 10),               # min_samples_split \n",
    "         (1, 10),               # min_samples_leaf \n",
    "         ('entropy', 'gini')]    # criterion\n",
    "\n",
    "# Name of the above features\n",
    "names = ['max_depth', 'max_features', 'min_samples_split', 'min_samples_leaf', 'criterion']\n",
    "\n",
    "# Number of iterations to perform\n",
    "n_calls = 15"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Optimize with SKOPT! Pass the objective function (*objective_func*) defined above and the *space* with parameter boundaries.\n",
    "\n",
    "Make use of **gp_minimize**, Bayesian optimization (sequential design strategy for global optimization of black-box functions) using Gaussian Processes (statistical model where observations occur in a continuous domain). The objective function is approximated by a Gaussian Process.\n",
    "\n",
    "Other minimization functions exists such as *forest_minimize* (Sequential optimisation using decision trees) or *gbrt_minimize* (Sequential optimization using gradient boosted trees). "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration No: 1 started. Evaluating function at random point.\n",
      "Iteration No: 1 ended. Evaluation done at random point.\n",
      "Time taken: 2.3323\n",
      "Function value obtained: -0.8498\n",
      "Current minimum: -0.8498\n",
      "Iteration No: 2 started. Evaluating function at random point.\n",
      "Iteration No: 2 ended. Evaluation done at random point.\n",
      "Time taken: 2.2157\n",
      "Function value obtained: -0.8474\n",
      "Current minimum: -0.8498\n",
      "Iteration No: 3 started. Evaluating function at random point.\n",
      "Iteration No: 3 ended. Evaluation done at random point.\n",
      "Time taken: 1.8051\n",
      "Function value obtained: -0.8444\n",
      "Current minimum: -0.8498\n",
      "Iteration No: 4 started. Evaluating function at random point.\n",
      "Iteration No: 4 ended. Evaluation done at random point.\n",
      "Time taken: 0.9917\n",
      "Function value obtained: -0.8526\n",
      "Current minimum: -0.8526\n",
      "Iteration No: 5 started. Evaluating function at random point.\n",
      "Iteration No: 5 ended. Evaluation done at random point.\n",
      "Time taken: 1.4009\n",
      "Function value obtained: -0.8539\n",
      "Current minimum: -0.8539\n",
      "Iteration No: 6 started. Evaluating function at random point.\n",
      "Iteration No: 6 ended. Evaluation done at random point.\n",
      "Time taken: 0.6734\n",
      "Function value obtained: -0.8484\n",
      "Current minimum: -0.8539\n",
      "Iteration No: 7 started. Evaluating function at random point.\n",
      "Iteration No: 7 ended. Evaluation done at random point.\n",
      "Time taken: 1.3157\n",
      "Function value obtained: -0.8406\n",
      "Current minimum: -0.8539\n",
      "Iteration No: 8 started. Evaluating function at random point.\n",
      "Iteration No: 8 ended. Evaluation done at random point.\n",
      "Time taken: 1.5037\n",
      "Function value obtained: -0.8502\n",
      "Current minimum: -0.8539\n",
      "Iteration No: 9 started. Evaluating function at random point.\n",
      "Iteration No: 9 ended. Evaluation done at random point.\n",
      "Time taken: 0.8843\n",
      "Function value obtained: -0.8562\n",
      "Current minimum: -0.8562\n",
      "Iteration No: 10 started. Evaluating function at random point.\n",
      "Iteration No: 10 ended. Evaluation done at random point.\n",
      "Time taken: 1.6242\n",
      "Function value obtained: -0.8437\n",
      "Current minimum: -0.8562\n",
      "Iteration No: 11 started. Searching for the next optimal point.\n",
      "Iteration No: 11 ended. Search finished for the next optimal point.\n",
      "Time taken: 0.8104\n",
      "Function value obtained: -0.8539\n",
      "Current minimum: -0.8562\n",
      "Iteration No: 12 started. Searching for the next optimal point.\n",
      "Iteration No: 12 ended. Search finished for the next optimal point.\n",
      "Time taken: 0.8051\n",
      "Function value obtained: -0.8460\n",
      "Current minimum: -0.8562\n",
      "Iteration No: 13 started. Searching for the next optimal point.\n",
      "Iteration No: 13 ended. Search finished for the next optimal point.\n",
      "Time taken: 0.7027\n",
      "Function value obtained: -0.8492\n",
      "Current minimum: -0.8562\n",
      "Iteration No: 14 started. Searching for the next optimal point.\n",
      "Iteration No: 14 ended. Search finished for the next optimal point.\n",
      "Time taken: 0.8055\n",
      "Function value obtained: -0.7184\n",
      "Current minimum: -0.8562\n",
      "Iteration No: 15 started. Searching for the next optimal point.\n",
      "Iteration No: 15 ended. Search finished for the next optimal point.\n",
      "Time taken: 2.2433\n",
      "Function value obtained: -0.8390\n",
      "Current minimum: -0.8562\n"
     ]
    }
   ],
   "source": [
    "results_skopt = gp_minimize(objective_func, \n",
    "                            space, \n",
    "                            verbose=True, \n",
    "                            n_calls=n_calls)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "What are the parameters that led to the best accuracy that was achieved?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'max_features': 2, 'min_samples_leaf': 5, 'max_depth': 10, 'min_samples_split': 7, 'criterion': 'gini'}\n"
     ]
    }
   ],
   "source": [
    "best_params = dict(zip(names, results_skopt.x))\n",
    "results['SKOPT'] = -results_skopt.fun\n",
    "print(best_params)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Plot the convergence over iterations:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1046531d0>"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAp4AAAGZCAYAAAAgmzG2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTOMAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x108dac668>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from skopt.plots import plot_convergence\n",
    "plt.rcParams[\"figure.figsize\"] = (10, 6)\n",
    "plot_convergence((\"gp_optimize\", results_skopt))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Comparing Results\n",
    "Let us compare all results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Default:                0.7907\n",
      "GridSearchCV:           0.8549 (32 iterations)\n",
      "RandomizedGridSearchCV: 0.8551 (20 iterations)\n",
      "SKOPT:                  0.8562 (15 iterations)\n"
     ]
    }
   ],
   "source": [
    "print('Default:                %.4f' % results['Default']) \n",
    "print('GridSearchCV:           %.4f (32 iterations)' % results['GridSearchCV']) \n",
    "print('RandomizedGridSearchCV: %.4f (20 iterations)' % results['RandomizedGridSearchCV']) \n",
    "print('SKOPT:                  %.4f (15 iterations)' % results['SKOPT']) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# To Sum Up\n",
    "- Tuning hyperparameters can (significantly) increase your machine learning model performance.\n",
    "- Scikit-learn provides basic hyperparameters search algorithms such as Grid-Search and Randomized Search\n",
    "- Scikit-Optimize is a more advanced algorithm for hyperparameter search (among other uses) and it generally results in equivalent or better solutions, but within less evaluations (less time consuming).\n",
    "- Recommendation: avoid creating your own hyperparameter optimization algorithms - build on top of existing ones"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.4.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}