02_Model_Building.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model Building and Optimization \n",
    "\n",
    "This notebook is a follow-up to the <code> 01_EDA </code> exploration. Our goal will be to find a model that best predicts medical appointment no-shows using the Aquarela Advanced Analytics dataset. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import all packages\n",
    "\n",
    "import pandas as pd \n",
    "import numpy as np \n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib.legend_handler import HandlerLine2D\n",
    "from collections import Counter\n",
    "\n",
    "from imblearn.over_sampling import SMOTE\n",
    "from imblearn.under_sampling import RandomUnderSampler\n",
    "from imblearn.pipeline import Pipeline\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.linear_model import SGDClassifier\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.metrics import roc_curve\n",
    "from sklearn.metrics import roc_auc_score\n",
    "from sklearn.metrics import precision_recall_curve\n",
    "from sklearn.metrics import f1_score\n",
    "from sklearn.metrics import auc\n",
    "from sklearn.metrics import confusion_matrix"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## The Goal\n",
    "\n",
    "In this exploration we will look at a few models that are well-suited to binary classification problems to optimize our prediction of medical appointment no-shows. \n",
    "\n",
    "We established that our dataset has a class imbalance (20% positive cases - no-shows vs. 80% negative cases) which means that we need to think carefully about how to measure the success of our models. We'll consider overall performance by looking at ROC curves. However, given that our ultimate goal is to reduce no-shows, we want to focus on optimizing recall more so than precision. After all, if we classify a few people who are going to show up to their appointments as no-showers (the false positives), they might receive an extra text or call, but ultimately still be unaffected. That said, if we have a high false negative rate, we'll be missing opportunities to intervene in appointments that would otherwise be missed. \n",
    "\n",
    "The models under consideration are as follow:\n",
    "\n",
    "* Logistic Regression – Baseline for the model\n",
    "* K-Nearest Neighbours - Effective when the number of samples is large (in this case, 110k), and interesting to experiment with because of the number of parameters that can be tuned\n",
    "* Random Forest – With the sparsity of the data, the sub-sampling could be effective and might control for overfitting\n",
    "\n",
    "We will start by dividing the data into training and test sets. We will then tackle the class imbalance through a combination of SMOTE and undersampling. Once this is complete, the training set will once again be split to provide a training and validation set. \n",
    "\n",
    "Once we've gauged the performance of the standard models, we will choose one to optimize. We will consider the most suitable hyperparameters through a combination of using validation curves and GridSearchCV, comparing the results of both methods. \n",
    "\n",
    "Finally, we will evaluate our model against the test set and calculate the confusion matrix for the results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import our data\n",
    "\n",
    "df = pd.read_csv('cleaned_data.csv', index_col=0)\n",
    "\n",
    "# Choose relevant features for model-building - see 01_EDA for methodology behind feature selection \n",
    "\n",
    "df_model = df[['No-show','Scholarship', 'Hipertension', 'Diabetes', 'Alcoholism', 'Handcap', 'SMS_received', 'Sched_to_App_Time', 'Num_NS']]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now that we have a subset of the data that we will use for model-building, let's split our dateframe into a train and test set. We will later split it again into train and validation sets, but first we need to deal with the class imbalance. Ultimately, we'll have the following splits:\n",
    "\n",
    "* <b> Train: </b> 60% (~66,000 samples)\n",
    "* <b> Validation: </b> 20% (~22,000 samples)\n",
    "* <b> Test: </b> 20% (~22,000 samples)\n",
    "<br>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Define target and features\n",
    "\n",
    "X = df_model.drop('No-show', axis=1)\n",
    "y = df_model['No-show'].values\n",
    "\n",
    "# Set a random state & scale the data\n",
    "\n",
    "r_state = 42 \n",
    "X = StandardScaler().fit_transform(X)\n",
    "\n",
    "# Train/test split \n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, random_state = r_state)\n",
    "\n",
    "# Plot our training set to see the imbalance between classes\n",
    "\n",
    "counter = Counter(y_train)\n",
    "\n",
    "fig = plt.figure()\n",
    "ax = plt.bar(counter.keys(), counter.values(), color=['#9AAB89', '#8FB8CA'])\n",
    "plt.xticks(ticks=[0,1], labels=['Shows','No-shows'])\n",
    "plt.ylim([0,75000])\n",
    "plt.text(0, 70000, s='68,187',horizontalalignment='center')\n",
    "plt.text(1, 19000, s='17,398',horizontalalignment='center')\n",
    "plt.ylabel('Number of Appointments')\n",
    "plt.title('Appointment Breakdown for Imbalanced Classes')\n",
    "\n",
    "fig.savefig(\"images/imbalanced_class.png\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "With the barplot we can see the class imbalance we currently have on our modelling data. We will balance the classes by first using SMOTE to oversample the minority class and then using random undersampling to reduce the number of examples for the majority class.\n",
    "\n",
    "In SMOTE, a random example from the minority class chosen, and k of the nearest neighbors for that example are found. A neighbour is randomly selected and a synthetic example is created at a randomly selected point between the two examples in feature space."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Counter({0: 68187, 1: 17398})\n",
      "Counter({0: 54548, 1: 27274})\n"
     ]
    }
   ],
   "source": [
    "# Print counter of imbalanced classes so we can compare our class shift later on\n",
    "\n",
    "print(counter)\n",
    "\n",
    "# Define pipeline\n",
    "\n",
    "over = SMOTE(sampling_strategy = 0.4, random_state = r_state)\n",
    "under = RandomUnderSampler(sampling_strategy= 0.5, random_state = r_state)\n",
    "pipeline = Pipeline(steps=[('o', over), ('u', under)])\n",
    "\n",
    "# Transform the dataset\n",
    "\n",
    "X_train, y_train = pipeline.fit_resample(X_train, y_train)\n",
    "\n",
    "# Summarize the new class distribution\n",
    "\n",
    "counter = Counter(y_train)\n",
    "print(counter)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Great, let's now plot how our new rebalanced dataset looks. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure()\n",
    "ax = plt.bar(counter.keys(), counter.values(), color=['#9AAB89', '#8FB8CA'])\n",
    "plt.xticks(ticks=[0,1], labels=['Shows','No-shows'])\n",
    "plt.ylim([0,60000])\n",
    "plt.text(0, 56500, s='54,548',horizontalalignment='center')\n",
    "plt.text(1, 29000, s='27,274',horizontalalignment='center')\n",
    "plt.ylabel('Number of Appointments')\n",
    "plt.title('Appointment Breakdown for Balanced Classes')\n",
    "\n",
    "fig.savefig(\"images/balanced_class.png\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Our ratio used of no-shows to shows used to be 1:4 and now we're closer to 1:2. Now that we've rebalanced our classes, we can split our training set again so that we have a training and validation set to use for our model exploration."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_val, y_train, y_val = train_test_split(X_train, y_train, test_size = 0.25, random_state = r_state)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now let's fit the three models we're considering onto our training data, comparing their performance with ROC and precision-recall curves."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "No Skill: ROC AUC=0.500\n",
      "Logistic Regression: ROC AUC=0.902\n",
      "KNN Classifier: ROC AUC=0.921\n",
      "Random Forest Classifier: ROC AUC=0.935\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Generate a no skill prediction (majority class)\n",
    "\n",
    "ns_prob = [0 for _ in range(len(y_val))]\n",
    "\n",
    "# Get prediction probability and predicted label for a classifier \n",
    "\n",
    "def get_prob_pred(classifier, X_train, y_train, X_val):\n",
    "    classifier.fit(X_train, y_train)\n",
    "    cl_prob = classifier.predict_proba(X_val)\n",
    "    cl_prob = cl_prob[:,1]\n",
    "    cl_y_pred = classifier.predict(X_val)\n",
    "    return cl_prob, cl_y_pred\n",
    "    \n",
    "lr_prob, lr_pred = get_prob_pred(LogisticRegression(), X_train, y_train, X_val)  \n",
    "knn_prob, knn_pred = get_prob_pred(KNeighborsClassifier(), X_train, y_train, X_val)    \n",
    "rf_prob, rf_pred = get_prob_pred(RandomForestClassifier(), X_train, y_train, X_val)    \n",
    "\n",
    "\n",
    "# Calculate scores\n",
    "\n",
    "ns_auc = roc_auc_score(y_val, ns_prob)\n",
    "lr_auc = roc_auc_score(y_val, lr_prob)\n",
    "knn_auc = roc_auc_score(y_val, knn_prob)\n",
    "rf_auc = roc_auc_score(y_val, rf_prob)\n",
    "\n",
    "#Summarize scores\n",
    "\n",
    "print('No Skill: ROC AUC=%.3f' % (ns_auc))\n",
    "print('Logistic Regression: ROC AUC=%.3f' % (lr_auc))\n",
    "print('KNN Classifier: ROC AUC=%.3f' % (knn_auc))\n",
    "print('Random Forest Classifier: ROC AUC=%.3f' % (rf_auc))\n",
    "\n",
    "# Calculate ROC curves\n",
    "\n",
    "ns_fpr, ns_tpr, threshold = roc_curve(y_val, ns_prob)\n",
    "lr_fpr, lr_tpr, threshold = roc_curve(y_val, lr_prob)\n",
    "knn_fpr, knn_tpr, threshold = roc_curve(y_val, knn_prob)\n",
    "rf_fpr, rf_tpr, threshold = roc_curve(y_val, rf_prob)\n",
    "\n",
    "# Plot the ROC curve for the model\n",
    "fig = plt.figure(figsize=[12,10])\n",
    "\n",
    "plt.plot(cmap=plt.cm.YlGn)\n",
    "plt.plot(ns_fpr, ns_tpr, linestyle='--', label='No Skill')\n",
    "plt.plot(lr_fpr, lr_tpr, marker='.', markersize=3,label='Logistic Regression')\n",
    "plt.plot(knn_fpr, knn_tpr, marker='.',markersize=3, label='KNN Classifier')\n",
    "plt.plot(rf_fpr, rf_tpr, marker='.', markersize=3,label='Random Forest Classifier')\n",
    "\n",
    "\n",
    "# Axis labels\n",
    "\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.ylabel('True Positive Rate')\n",
    "\n",
    "# Show the legend\n",
    "\n",
    "plt.legend()\n",
    "\n",
    "# Show the plot\n",
    "\n",
    "fig.savefig(\"images/ROC_AUC.png\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From the ROC AUC score, it looks like the Random Forest Classifier is our most promising model to tune given its score of 0.935. However, before we confirm this, we need to have a look at the precision-recall curves. Remember, we are going to lean in the direction of models that have better recall, since we are willing to have more false negatives in favour of being able to capture more true positives (and as a result, we're willing to let in more false positives into our model. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Logistic Regression: recall=0.619 auc=0.731\n",
      "KNN Classifier: recall=0.816 auc=0.818\n",
      "Random Forest Classifier: recall=0.833 auc=0.824\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Predict precision and recall\n",
    "\n",
    "lr_precision, lr_recall, threshold = precision_recall_curve(y_val, lr_prob)\n",
    "lr_f1, lr_auc = f1_score(y_val, lr_pred), auc(lr_recall, lr_precision)\n",
    "\n",
    "knn_precision, knn_recall, threshold = precision_recall_curve(y_val, knn_prob)\n",
    "knn_f1, knn_auc = f1_score(y_val, knn_pred), auc(knn_recall, knn_precision)\n",
    "\n",
    "rf_precision, rf_recall, threshold = precision_recall_curve(y_val, rf_prob)\n",
    "rf_f1, rf_auc = f1_score(y_val, rf_pred), auc(rf_recall, rf_precision)\n",
    "\n",
    "# Summarize scores\n",
    "\n",
    "print('Logistic Regression: recall=%.3f auc=%.3f' % (lr_f1, lr_auc))\n",
    "print('KNN Classifier: recall=%.3f auc=%.3f' % (knn_f1, knn_auc))\n",
    "print('Random Forest Classifier: recall=%.3f auc=%.3f' % (rf_f1, rf_auc))\n",
    "\n",
    "# Plot the precision-recall curves\n",
    "\n",
    "fig = plt.figure(figsize=[12,10])\n",
    "\n",
    "no_skill = len(y_val[y_val==1]) / len(y_val)\n",
    "plt.plot([0, 1], [no_skill, no_skill], linestyle='--', label='No Skill')\n",
    "plt.plot(lr_recall, lr_precision, marker='.', markersize=3, label='Logistic Regression')\n",
    "plt.plot(knn_recall, knn_precision, marker='.', markersize=3, label='KNN Classifier')\n",
    "plt.plot(rf_recall, rf_precision, marker='.', markersize=3, label='Random Forest Classifier')\n",
    "\n",
    "# Axis labels\n",
    "\n",
    "plt.xlabel('Recall')\n",
    "plt.ylabel('Precision')\n",
    "\n",
    "# Show the legend\n",
    "\n",
    "plt.legend()\n",
    "\n",
    "# Show the plot\n",
    "\n",
    "fig.savefig(\"images/recall.png\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It looks like our Random Forest Classifier is also the best performing in terms of recall. Considering both the ROC and Recall of this classifier outperform the others we've examined, let's fine-tune a few hyperparameters of this model and see how it performs against our validation set.\n",
    "\n",
    "Once we've done this, we can retrain our tuned model on the training and validation data combined, and then give it a final evaluation on our test set. \n",
    "\n",
    "The hyperparameters we'll consider are as follow:\n",
    "\n",
    "* <b>n_estimators:</b> This specifies the number of trees in the model's forest - default value is 10. \n",
    "* <b>max_depth:</b> This specifies the maximum depth of each tree - default value is 'None', which means the nodes are expanded until all leaves are pure or until all leaves contain less than min_samples_split samples.\n",
    "* <b>min_samples_split:</b> he minimum number of samples required to split an internal node - default is 2, which means an internal node must have at least two samples to be split into a more specific classification.\n",
    "* <b>min_samples_leaf:</b> The minimum number of samples required to be at a leaf node - default is 1, which means every leaf hs to have at least one sample to classified. \n",
    "\n",
    "For each hyperparameter, let's evaluate the ROC AUC performance on both the training and validation data. We begin our exploration with 'n_estimators'."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "64 0.8881440966537801\n"
     ]
    }
   ],
   "source": [
    "# Define a list of hyperparameter values we want to consider \n",
    "\n",
    "n_estimators = [1, 2, 4, 8, 16, 32, 64, 100, 200]\n",
    "\n",
    "# Empty lists to hold our training and validation results for the ROC AUC curves\n",
    "\n",
    "train_results = []\n",
    "val_results = []\n",
    "\n",
    "# Loop over the range of hyperparameter values, calculate ROC AUC for both the\n",
    "# training and validation data, and append them to our lists\n",
    "\n",
    "for estimator in n_estimators:\n",
    "   rf = RandomForestClassifier(n_estimators=estimator, n_jobs=-1)\n",
    "   rf.fit(X_train, y_train)\n",
    "   train_pred = rf.predict(X_train)\n",
    "   fpr, tpr, thresholds = roc_curve(y_train, train_pred)\n",
    "   roc_auc = auc(fpr, tpr)\n",
    "   train_results.append(roc_auc)\n",
    "   val_pred = rf.predict(X_val)\n",
    "   fpr, tpr, thresholds = roc_curve(y_val, val_pred)\n",
    "   roc_auc = auc(fpr, tpr)\n",
    "   val_results.append(roc_auc)\n",
    "\n",
    "# Plot the results \n",
    "fig = plt.figure()\n",
    "line1, = plt.plot(n_estimators, train_results, 'b', label='Train AUC')\n",
    "line2, = plt.plot(n_estimators, val_results, 'r', label='Validation AUC')\n",
    "\n",
    "plt.legend(handler_map={line1: HandlerLine2D(numpoints=2)})\n",
    "plt.ylabel('AUC score')\n",
    "plt.xlabel('n_estimators')\n",
    "\n",
    "fig.savefig(\"images/n_estimators.png\")\n",
    "plt.show()\n",
    "\n",
    "# Get the optimum AUC score\n",
    "\n",
    "max_y = max(val_results)  # Find the maximum y value\n",
    "max_x = n_estimators[val_results.index(max_y)]  # Find the x value corresponding to the maximum y value\n",
    "print (max_x, max_y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can see that our model sharply increases its AUC score and then tapers off. With that in mind, the best number of estimators for us to use in our final model is 64. Let's move on to examining max_depth."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "9.0 0.8965512651121598\n"
     ]
    }
   ],
   "source": [
    "# Define a list of hyperparameter values we want to consider \n",
    "\n",
    "max_depths = np.linspace(1, 32, 32, endpoint=True)\n",
    "\n",
    "# Empty lists to hold our training and validation results for the ROC AUC curves\n",
    "\n",
    "train_results = []\n",
    "val_results = []\n",
    "\n",
    "# Loop over the range of hyperparameter values, calculate ROC AUC for both the\n",
    "# training and validation data, and append them to our lists\n",
    "\n",
    "for max_depth in max_depths:\n",
    "   rf = RandomForestClassifier(n_estimators = 64, max_depth=max_depth, n_jobs=-1)\n",
    "   rf.fit(X_train, y_train)\n",
    "   train_pred = rf.predict(X_train)\n",
    "   fpr, tpr, thresholds = roc_curve(y_train, train_pred)\n",
    "   roc_auc = auc(fpr, tpr)\n",
    "   train_results.append(roc_auc)\n",
    "   val_pred = rf.predict(X_val)\n",
    "   fpr, tpr, thresholds = roc_curve(y_val, val_pred)\n",
    "   roc_auc = auc(fpr, tpr)\n",
    "   val_results.append(roc_auc)\n",
    "    \n",
    "# Plot the results \n",
    "\n",
    "fig= plt.figure()\n",
    "line1, = plt.plot(max_depths, train_results, 'b', label='Train AUC')\n",
    "line2, = plt.plot(max_depths, val_results, 'r', label='Validation AUC')\n",
    "\n",
    "plt.legend(handler_map={line1: HandlerLine2D(numpoints=2)})\n",
    "plt.ylabel('AUC score')\n",
    "plt.xlabel('max_depth')\n",
    "\n",
    "fig.savefig(\"images/max_depth.png\")\n",
    "plt.show()\n",
    "\n",
    "# Get the optimum AUC score\n",
    "\n",
    "max_y = max(val_results)  # Find the maximum y value\n",
    "max_x = max_depths[val_results.index(max_y)]  # Find the x value corresponding to the maximum y value\n",
    "print (max_x, max_y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Looks like a depth of 8 gives us the best results for depth. Next up is the minimum sample split. We'll use a similar technique with this hyperparameter."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.1 0.8908441672335159\n"
     ]
    }
   ],
   "source": [
    "# Define a list of hyperparameter values we want to consider \n",
    "\n",
    "min_samples_splits = np.linspace(0.1, 1.0, 10, endpoint=True)\n",
    "\n",
    "# Empty lists to hold our training and validation results for the ROC AUC curves\n",
    "\n",
    "train_results = []\n",
    "val_results = []\n",
    "\n",
    "# Loop over the range of hyperparameter values, calculate ROC AUC for both the\n",
    "# training and validation data, and append them to our lists\n",
    "\n",
    "for min_samples_split in min_samples_splits:\n",
    "   rf = RandomForestClassifier(n_estimators = 64, max_depth=8, min_samples_split=min_samples_split, n_jobs=-1)\n",
    "   rf.fit(X_train, y_train)\n",
    "   train_pred = rf.predict(X_train)\n",
    "   fpr, tpr, thresholds = roc_curve(y_train, train_pred)\n",
    "   roc_auc = auc(fpr, tpr)\n",
    "   train_results.append(roc_auc)\n",
    "   val_pred = rf.predict(X_val)\n",
    "   fpr, tpr, thresholds = roc_curve(y_val, val_pred)\n",
    "   roc_auc = auc(fpr, tpr)\n",
    "   val_results.append(roc_auc)\n",
    "    \n",
    "# Plot the results \n",
    "\n",
    "fig= plt.figure()\n",
    "line1, = plt.plot(min_samples_splits, train_results, 'b', label='Train AUC')\n",
    "line2, = plt.plot(min_samples_splits, val_results, 'r', label='Validation AUC')\n",
    "\n",
    "plt.legend(handler_map={line1: HandlerLine2D(numpoints=2)})\n",
    "plt.ylabel('AUC score')\n",
    "plt.xlabel('max_depth')\n",
    "\n",
    "fig.savefig(\"images/min_sample_splits.png\")\n",
    "plt.show()\n",
    "\n",
    "# Get the optimum AUC score\n",
    "\n",
    "max_y = max(val_results)  # Find the maximum y value\n",
    "max_x = min_samples_splits[val_results.index(max_y)]  # Find the x value corresponding to the maximum y value\n",
    "print (max_x, max_y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Our optimal min_samples_split is 0.1, giving us an ROC AUC score of 0.89. Finally, let's consider our min_samples_leaf parameter."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "67.0 0.8911019927766612\n"
     ]
    }
   ],
   "source": [
    "# Define a list of hyperparameter values we want to consider \n",
    "\n",
    "min_samples_leafs = np.linspace(1,100,10, endpoint=True)\n",
    "\n",
    "# Empty lists to hold our training and validation results for the ROC AUC curves\n",
    "\n",
    "train_results = []\n",
    "val_results = []\n",
    "\n",
    "# Loop over the range of hyperparameter values, calculate ROC AUC for both the\n",
    "# training and validation data, and append them to our lists\n",
    "\n",
    "for min_samples_leaf in min_samples_leafs:\n",
    "   rf = RandomForestClassifier(n_estimators = 64, max_depth=8,min_samples_split= 0.1,min_samples_leaf=int(min_samples_leaf), n_jobs=-1)\n",
    "   rf.fit(X_train, y_train)\n",
    "   train_pred = rf.predict(X_train)\n",
    "   fpr, tpr, thresholds = roc_curve(y_train, train_pred)\n",
    "   roc_auc = auc(fpr, tpr)\n",
    "   train_results.append(roc_auc)\n",
    "   val_pred = rf.predict(X_val)\n",
    "   fpr, tpr, thresholds = roc_curve(y_val, val_pred)\n",
    "   roc_auc = auc(fpr, tpr)\n",
    "   val_results.append(roc_auc)\n",
    "    \n",
    "# Plot the results \n",
    "\n",
    "fig = plt.figure()\n",
    "line1, = plt.plot(min_samples_leafs, train_results, 'b', label='Train AUC')\n",
    "line2, = plt.plot(min_samples_leafs, val_results, 'r', label='Validation AUC')\n",
    "\n",
    "plt.legend(handler_map={line1: HandlerLine2D(numpoints=2)})\n",
    "plt.ylabel('AUC score')\n",
    "plt.xlabel('max_depth')\n",
    "\n",
    "fig.savefig(\"images/min_samples_leaf.png\")\n",
    "plt.show()\n",
    "\n",
    "# Get the optimum AUC score\n",
    "\n",
    "max_y = max(val_results)  # Find the maximum y value\n",
    "max_x = min_samples_leafs[val_results.index(max_y)]  # Find the x value corresponding to the maximum y value\n",
    "print (max_x, max_y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Our best min_samples_leaf value is 67. From our manual exploration, we've gotten the following optimal values:\n",
    "\n",
    "* <b>n_estimators:</b> 64\n",
    "* <b>max_depth:</b> 8\n",
    "* <b>min_samples_split:</b> 0.1\n",
    "* <b>min_samples_leaf:</b> 67\n",
    "\n",
    "Let's now compare these results with a GridSearch - the advantage of a GridSearch is that it allows for permutations of different combinations of hyperparameters more easily than manual exploration. We will see whether this improves the overall performance of our model. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 305,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 31680 candidates, totalling 95040 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:    5.4s\n",
      "[Parallel(n_jobs=-1)]: Done 184 tasks      | elapsed:   22.1s\n",
      "[Parallel(n_jobs=-1)]: Done 434 tasks      | elapsed:   46.7s\n",
      "[Parallel(n_jobs=-1)]: Done 784 tasks      | elapsed:  1.3min\n",
      "[Parallel(n_jobs=-1)]: Done 1234 tasks      | elapsed:  2.0min\n",
      "[Parallel(n_jobs=-1)]: Done 1784 tasks      | elapsed:  2.9min\n",
      "[Parallel(n_jobs=-1)]: Done 2434 tasks      | elapsed:  3.9min\n",
      "[Parallel(n_jobs=-1)]: Done 3184 tasks      | elapsed:  5.2min\n",
      "[Parallel(n_jobs=-1)]: Done 4034 tasks      | elapsed:  6.7min\n",
      "[Parallel(n_jobs=-1)]: Done 4984 tasks      | elapsed:  8.4min\n",
      "[Parallel(n_jobs=-1)]: Done 6034 tasks      | elapsed: 10.2min\n",
      "[Parallel(n_jobs=-1)]: Done 7184 tasks      | elapsed: 12.2min\n",
      "[Parallel(n_jobs=-1)]: Done 8434 tasks      | elapsed: 14.3min\n",
      "[Parallel(n_jobs=-1)]: Done 9784 tasks      | elapsed: 16.7min\n",
      "[Parallel(n_jobs=-1)]: Done 11234 tasks      | elapsed: 19.3min\n",
      "[Parallel(n_jobs=-1)]: Done 12784 tasks      | elapsed: 21.9min\n",
      "[Parallel(n_jobs=-1)]: Done 14434 tasks      | elapsed: 24.8min\n",
      "[Parallel(n_jobs=-1)]: Done 16184 tasks      | elapsed: 27.8min\n",
      "[Parallel(n_jobs=-1)]: Done 18034 tasks      | elapsed: 31.1min\n",
      "[Parallel(n_jobs=-1)]: Done 19984 tasks      | elapsed: 34.5min\n",
      "[Parallel(n_jobs=-1)]: Done 22034 tasks      | elapsed: 38.1min\n",
      "[Parallel(n_jobs=-1)]: Done 24184 tasks      | elapsed: 41.9min\n",
      "[Parallel(n_jobs=-1)]: Done 26434 tasks      | elapsed: 45.8min\n",
      "[Parallel(n_jobs=-1)]: Done 28784 tasks      | elapsed: 49.9min\n",
      "[Parallel(n_jobs=-1)]: Done 31234 tasks      | elapsed: 54.2min\n",
      "[Parallel(n_jobs=-1)]: Done 33784 tasks      | elapsed: 58.7min\n",
      "[Parallel(n_jobs=-1)]: Done 36434 tasks      | elapsed: 63.4min\n",
      "[Parallel(n_jobs=-1)]: Done 39184 tasks      | elapsed: 68.3min\n",
      "[Parallel(n_jobs=-1)]: Done 42034 tasks      | elapsed: 73.4min\n",
      "[Parallel(n_jobs=-1)]: Done 44984 tasks      | elapsed: 78.6min\n",
      "[Parallel(n_jobs=-1)]: Done 48034 tasks      | elapsed: 84.1min\n",
      "[Parallel(n_jobs=-1)]: Done 51184 tasks      | elapsed: 89.7min\n",
      "[Parallel(n_jobs=-1)]: Done 54434 tasks      | elapsed: 95.6min\n",
      "[Parallel(n_jobs=-1)]: Done 57784 tasks      | elapsed: 101.7min\n",
      "[Parallel(n_jobs=-1)]: Done 61234 tasks      | elapsed: 108.2min\n",
      "[Parallel(n_jobs=-1)]: Done 64784 tasks      | elapsed: 114.8min\n",
      "[Parallel(n_jobs=-1)]: Done 68434 tasks      | elapsed: 121.6min\n",
      "[Parallel(n_jobs=-1)]: Done 72184 tasks      | elapsed: 128.5min\n",
      "[Parallel(n_jobs=-1)]: Done 76034 tasks      | elapsed: 135.5min\n",
      "[Parallel(n_jobs=-1)]: Done 79984 tasks      | elapsed: 142.6min\n",
      "[Parallel(n_jobs=-1)]: Done 84034 tasks      | elapsed: 149.9min\n",
      "[Parallel(n_jobs=-1)]: Done 88184 tasks      | elapsed: 157.4min\n",
      "[Parallel(n_jobs=-1)]: Done 92434 tasks      | elapsed: 165.1min\n",
      "[Parallel(n_jobs=-1)]: Done 95040 out of 95040 | elapsed: 169.8min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9333500798171329\n",
      "RandomForestClassifier(max_depth=3.0, min_samples_split=0.1, n_estimators=4,\n",
      "                       random_state=42)\n"
     ]
    }
   ],
   "source": [
    "# Outline the hyperparameter values we want to consider\n",
    "\n",
    "n_estimators = [1, 2, 4, 8, 16, 32, 64, 100, 200]\n",
    "max_depth = np.linspace(1, 32, 32, endpoint=True)\n",
    "min_samples_split = np.linspace(0.1, 1.0, 10, endpoint=True)\n",
    "min_samples_leaf = [1,10,20,30,40,50,60,70,80,90,100]\n",
    "\n",
    "# Combine into a dictionary\n",
    "\n",
    "parameters = dict(n_estimators = n_estimators, max_depth = max_depth,  \n",
    "              min_samples_split = min_samples_split, \n",
    "             min_samples_leaf = min_samples_leaf)\n",
    "\n",
    "# Instantiate GridSearch\n",
    "\n",
    "rf_gs = RandomForestClassifier(random_state=r_state)\n",
    "\n",
    "gs_rf = GridSearchCV(rf_gs, parameters, cv = 3, verbose = 1, \n",
    "                      n_jobs = -1, scoring='roc_auc')\n",
    "\n",
    "best_gs = gs_rf.fit(X_train, y_train)\n",
    "\n",
    "# Get best parameters and best score\n",
    "\n",
    "print(best_gs.best_score_)\n",
    "print(best_gs.best_estimator_)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We've gotten our ideal parameters and the ROC AUC score they lead to on the training set - it's a bit lower than what we had with the default hyperparameters, but this is likely because the cross validation on Grid Search gave us the mean of 3 cross-validations, so while our performance is a bit weaker, it suffers from less ovrefitting, and will thus be more helpful for the validation data. Let's test  the default random forest, our manually tuned random forest, and our grid search random forest on the validation data to compare performance."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Default Hyperparameters: recall=0.834 ROC AUC=0.935\n",
      "Manual Hyperparameters: recall=0.826 ROC AUC=0.930\n",
      "GridSearch Hyperparameters: recall=0.836 ROC AUC=0.933\n"
     ]
    }
   ],
   "source": [
    "# Define our 3 classifiers \n",
    "\n",
    "rf = RandomForestClassifier(random_state=r_state)\n",
    "man_rf = RandomForestClassifier(n_estimators = 64, max_depth= 8, min_samples_split = 0.1, min_samples_leaf = 67, random_state=r_state)\n",
    "grid_rf = RandomForestClassifier(n_estimators = 4, max_depth=3, min_samples_split=0.1,min_samples_leaf = 13,random_state=r_state)\n",
    "\n",
    "# Fit them to the training data and get predictions\n",
    "\n",
    "rf_prob, rf_pred = get_prob_pred(rf, X_train, y_train, X_val)\n",
    "manrf_prob, manrf_pred = get_prob_pred(man_rf, X_train, y_train, X_val)\n",
    "gridrf_prob, gridrf_pred = get_prob_pred(grid_rf, X_train, y_train, X_val)\n",
    "\n",
    "# Get their ROC and recall scores for the validation set\n",
    "\n",
    "rf_vauc = roc_auc_score(y_val, rf_prob)\n",
    "\n",
    "rf_precision, rf_recall, threshold = precision_recall_curve(y_val, rf_prob)\n",
    "rf_f1, rf_auc = f1_score(y_val, rf_pred), auc(rf_recall, rf_precision)\n",
    "\n",
    "manrf_vauc = roc_auc_score(y_val, manrf_prob)\n",
    "\n",
    "manrf_precision, manrf_recall, threshold = precision_recall_curve(y_val, manrf_prob)\n",
    "manrf_f1, manrf_auc = f1_score(y_val, manrf_pred), auc(manrf_recall, manrf_precision)\n",
    "\n",
    "gridrf_vauc = roc_auc_score(y_val, gridrf_prob)\n",
    "\n",
    "gridrf_precision, gridrf_recall, threshold = precision_recall_curve(y_val, gridrf_prob)\n",
    "gridrf_f1, gridrf_auc = f1_score(y_val, gridrf_pred), auc(gridrf_recall, gridrf_precision)\n",
    "\n",
    "# Summarize scores\n",
    "\n",
    "print('Default Hyperparameters: recall=%.3f ROC AUC=%.3f' % (rf_f1, rf_vauc))\n",
    "print('Manual Hyperparameters: recall=%.3f ROC AUC=%.3f' % (manrf_f1, manrf_vauc))\n",
    "print('GridSearch Hyperparameters: recall=%.3f ROC AUC=%.3f' % (gridrf_f1, gridrf_vauc))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can see that the GridSearch gives us marginally better recall than the default parameters, and ultimately that's the priority for this investigation. With that in mind, let's recombine our test and validation data, retrain our model, and use the GridSearch version of our RF classifier to gauge performance against the test set. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_final = np.concatenate((X_train, X_val), axis=0)\n",
    "y_train_final = np.concatenate((y_train, y_val), axis=0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now let's retrain our model and see its performance against the test set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Final model performance: recall=0.740 ROC AUC=0.933\n"
     ]
    }
   ],
   "source": [
    "# Fit to the training data and get predictions\n",
    "\n",
    "gridrf_prob, gridrf_pred = get_prob_pred(grid_rf, X_train_final, y_train_final, X_test)\n",
    "\n",
    "# Get ROC and recall scores for the test set\n",
    "\n",
    "gridrf_rauc = roc_auc_score(y_test, gridrf_prob)\n",
    "\n",
    "gridrf_precision, gridrf_recall, threshold = precision_recall_curve(y_test, gridrf_prob)\n",
    "gridrf_f1, gridrf_auc = f1_score(y_test, gridrf_pred), auc(gridrf_recall, gridrf_precision)\n",
    "\n",
    "# Summarize scores\n",
    "\n",
    "print('Final model performance: recall=%.3f ROC AUC=%.3f' % (gridrf_f1, gridrf_rauc))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Our recall score has dropped quite a bit, but we still have a pretty robust ROC AUC with this model. Let's have a look at the confusion matrix corresponding to our model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "cmat_rf = confusion_matrix(y_test, gridrf_pred)\n",
    "\n",
    "fig = plt.figure()\n",
    "def heatmap_labels(x):\n",
    "    group_names = ['True Neg','False Pos','False Neg','True Pos']\n",
    "    group_counts = [\"{0:0.0f}\".format(value) for value in\n",
    "                x.flatten()]\n",
    "    group_percentages = [\"{0:.2%}\".format(value) for value in\n",
    "                     x.flatten()/np.sum(x)]\n",
    "    labels = [f\"{v1}\\n{v2}\\n{v3}\"for v1, v2, v3 in\n",
    "          zip(group_names,group_counts,group_percentages)]\n",
    "    labels = np.asarray(labels).reshape(2,2)\n",
    "    return sns.heatmap(x, annot=labels, fmt='', cmap='Blues')\n",
    "\n",
    "heatmap_labels(cmat_rf)\n",
    "\n",
    "fig.savefig(\"images/final_model_heatmap.png\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This is a great result for our purposes! Recall that our purpose was to identify as many true positives as possible, and we were willing to accept a higher number of false positives in exchange because the interventions we would implement to mitigate medical appointment no-shows likely wouldn't deter those who were intending to show up to appointments but received a little more 'attention' because they were falsely marked as likely to no-show. \n",
    "\n",
    "Given that this model can identify 91% of our true positive cases, it can be very useful for healthcare practitioners when targeting patients for interventions. "
   ]
  }
 ],
 "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": "3.8.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}