kNN_stock_vs_tuned.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Goal:** Compare the base model's performance for the KNeighborsClassifier (kNN) to the tuned model's performance using a cleaned UCI-ML Repo dataset (Speech Recognition)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# necessary imports for EDA\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "%matplotlib inline\n",
    "\n",
    "# sklearn imports\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.model_selection import cross_val_score, train_test_split, GridSearchCV\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.metrics import classification_report, confusion_matrix, plot_confusion_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "The dataset is quite small. There are  329  rows and  13 columns.\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>language</th>\n",
       "      <th>X1</th>\n",
       "      <th>X2</th>\n",
       "      <th>X3</th>\n",
       "      <th>X4</th>\n",
       "      <th>X5</th>\n",
       "      <th>X6</th>\n",
       "      <th>X7</th>\n",
       "      <th>X8</th>\n",
       "      <th>X9</th>\n",
       "      <th>X10</th>\n",
       "      <th>X11</th>\n",
       "      <th>X12</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ES</td>\n",
       "      <td>7.071476</td>\n",
       "      <td>-6.512900</td>\n",
       "      <td>7.650800</td>\n",
       "      <td>11.150783</td>\n",
       "      <td>-7.657312</td>\n",
       "      <td>12.484021</td>\n",
       "      <td>-11.709772</td>\n",
       "      <td>3.426596</td>\n",
       "      <td>1.462715</td>\n",
       "      <td>-2.812753</td>\n",
       "      <td>0.866538</td>\n",
       "      <td>-5.244274</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ES</td>\n",
       "      <td>10.982967</td>\n",
       "      <td>-5.157445</td>\n",
       "      <td>3.952060</td>\n",
       "      <td>11.529381</td>\n",
       "      <td>-7.638047</td>\n",
       "      <td>12.136098</td>\n",
       "      <td>-12.036247</td>\n",
       "      <td>3.491943</td>\n",
       "      <td>0.595441</td>\n",
       "      <td>-4.508811</td>\n",
       "      <td>2.332147</td>\n",
       "      <td>-6.221857</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>ES</td>\n",
       "      <td>7.827108</td>\n",
       "      <td>-5.477472</td>\n",
       "      <td>7.816257</td>\n",
       "      <td>9.187592</td>\n",
       "      <td>-7.172511</td>\n",
       "      <td>11.715299</td>\n",
       "      <td>-13.847214</td>\n",
       "      <td>4.574075</td>\n",
       "      <td>-1.687559</td>\n",
       "      <td>-7.204041</td>\n",
       "      <td>-0.011847</td>\n",
       "      <td>-6.463144</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ES</td>\n",
       "      <td>6.744083</td>\n",
       "      <td>-5.688920</td>\n",
       "      <td>6.546789</td>\n",
       "      <td>9.000183</td>\n",
       "      <td>-6.924963</td>\n",
       "      <td>11.710766</td>\n",
       "      <td>-12.374388</td>\n",
       "      <td>6.169879</td>\n",
       "      <td>-0.544747</td>\n",
       "      <td>-6.019237</td>\n",
       "      <td>1.358559</td>\n",
       "      <td>-6.356441</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ES</td>\n",
       "      <td>5.836843</td>\n",
       "      <td>-5.326557</td>\n",
       "      <td>7.472265</td>\n",
       "      <td>8.847440</td>\n",
       "      <td>-6.773244</td>\n",
       "      <td>12.677218</td>\n",
       "      <td>-12.315061</td>\n",
       "      <td>4.416344</td>\n",
       "      <td>0.193500</td>\n",
       "      <td>-3.644812</td>\n",
       "      <td>2.151239</td>\n",
       "      <td>-6.816310</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  language         X1        X2        X3         X4        X5         X6  \\\n",
       "0       ES   7.071476 -6.512900  7.650800  11.150783 -7.657312  12.484021   \n",
       "1       ES  10.982967 -5.157445  3.952060  11.529381 -7.638047  12.136098   \n",
       "2       ES   7.827108 -5.477472  7.816257   9.187592 -7.172511  11.715299   \n",
       "3       ES   6.744083 -5.688920  6.546789   9.000183 -6.924963  11.710766   \n",
       "4       ES   5.836843 -5.326557  7.472265   8.847440 -6.773244  12.677218   \n",
       "\n",
       "          X7        X8        X9       X10       X11       X12  \n",
       "0 -11.709772  3.426596  1.462715 -2.812753  0.866538 -5.244274  \n",
       "1 -12.036247  3.491943  0.595441 -4.508811  2.332147 -6.221857  \n",
       "2 -13.847214  4.574075 -1.687559 -7.204041 -0.011847 -6.463144  \n",
       "3 -12.374388  6.169879 -0.544747 -6.019237  1.358559 -6.356441  \n",
       "4 -12.315061  4.416344  0.193500 -3.644812  2.151239 -6.816310  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 329 entries, 0 to 328\n",
      "Data columns (total 13 columns):\n",
      " #   Column    Non-Null Count  Dtype  \n",
      "---  ------    --------------  -----  \n",
      " 0   language  329 non-null    object \n",
      " 1   X1        329 non-null    float64\n",
      " 2   X2        329 non-null    float64\n",
      " 3   X3        329 non-null    float64\n",
      " 4   X4        329 non-null    float64\n",
      " 5   X5        329 non-null    float64\n",
      " 6   X6        329 non-null    float64\n",
      " 7   X7        329 non-null    float64\n",
      " 8   X8        329 non-null    float64\n",
      " 9   X9        329 non-null    float64\n",
      " 10  X10       329 non-null    float64\n",
      " 11  X11       329 non-null    float64\n",
      " 12  X12       329 non-null    float64\n",
      "dtypes: float64(12), object(1)\n",
      "memory usage: 33.5+ KB\n"
     ]
    }
   ],
   "source": [
    "# load data\n",
    "srd = pd.read_csv('../Machine-Learning-Data/accent-mfcc-data-1.csv')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "srd.info()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Cleaning**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# convert the target variable to categorical\n",
    "srd['language'] = srd['language'].astype('category')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**EDA**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "#sns.pairplot(srd, hue = 'language')\n",
    "#plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Model Set-up**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create train and test sets\n",
    "X = srd.drop(['language'], axis = 'columns')\n",
    "y = srd['language']\n",
    "\n",
    "# random sampling\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = .25, random_state = 28)\n",
    "\n",
    "# stratified sampling\n",
    "X_train_s, X_test_s, y_train_s, y_test_s = train_test_split(X, y, test_size = .25, random_state = 28, stratify = y)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>y_train</th>\n",
       "      <th>y_test</th>\n",
       "      <th>y_train_s</th>\n",
       "      <th>y_test_s</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>US</th>\n",
       "      <td>0.487805</td>\n",
       "      <td>0.542169</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>0.506024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>UK</th>\n",
       "      <td>0.150407</td>\n",
       "      <td>0.096386</td>\n",
       "      <td>0.138211</td>\n",
       "      <td>0.132530</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>FR</th>\n",
       "      <td>0.101626</td>\n",
       "      <td>0.060241</td>\n",
       "      <td>0.093496</td>\n",
       "      <td>0.084337</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>IT</th>\n",
       "      <td>0.097561</td>\n",
       "      <td>0.072289</td>\n",
       "      <td>0.089431</td>\n",
       "      <td>0.096386</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ES</th>\n",
       "      <td>0.093496</td>\n",
       "      <td>0.072289</td>\n",
       "      <td>0.089431</td>\n",
       "      <td>0.084337</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>GE</th>\n",
       "      <td>0.069106</td>\n",
       "      <td>0.156627</td>\n",
       "      <td>0.089431</td>\n",
       "      <td>0.096386</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     y_train    y_test  y_train_s  y_test_s\n",
       "US  0.487805  0.542169   0.500000  0.506024\n",
       "UK  0.150407  0.096386   0.138211  0.132530\n",
       "FR  0.101626  0.060241   0.093496  0.084337\n",
       "IT  0.097561  0.072289   0.089431  0.096386\n",
       "ES  0.093496  0.072289   0.089431  0.084337\n",
       "GE  0.069106  0.156627   0.089431  0.096386"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# check proportions of each category for both samples\n",
    "prop_counts = pd.DataFrame()\n",
    "prop_counts['y_train'] = pd.Series(y_train.value_counts(normalize = True))\n",
    "prop_counts['y_test'] = pd.Series(y_test.value_counts(normalize = True))\n",
    "prop_counts['y_train_s'] = pd.Series(y_train_s.value_counts(normalize = True))\n",
    "prop_counts['y_test_s'] = pd.Series(y_test_s.value_counts(normalize = True))\n",
    "display(prop_counts)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The data frame shows the stratified train/test split has similar proportions via category. This might actually hurt the model as there are fewer samples to further split during cross-validation. My assumption is that I will have to use 3-5 fold cross-validation to achieve higher accuracy for the stratified split, while I can use 5-10 fold cross-validation for the random split."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Stock Model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "KNeighborsClassifier()"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# create the classifier and leave defaults values: p = 2 (Euclidean Distance), weights = 'uniform', n_neighbors = 5\n",
    "knn_base = KNeighborsClassifier()\n",
    "knn_base_s = KNeighborsClassifier()\n",
    "\n",
    "# fit the model to both samples\n",
    "knn_base.fit(X_train, y_train)\n",
    "knn_base_s.fit(X_train_s, y_train_s)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Check the overall classification accuracy using the score method. It appears that the random sample produces slightly better accuracy than the stratified sample for the stock model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The overall accuracy is:  80.0 %\n",
      "The overall accuracy is:  77.0 %\n"
     ]
    }
   ],
   "source": [
    "# random\n",
    "print('The overall accuracy is: ', round(knn_base.score(X_test, y_test),2) * 100, '%')\n",
    "# stratified\n",
    "print('The overall accuracy is: ', round(knn_base_s.score(X_test_s, y_test_s),2) * 100, '%')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Random Overall Accuracy:  80 %\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Stratified Overall Accuracy:  77 %\n"
     ]
    },
    {
     "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": [
    "# make the predictions and check the score method manually\n",
    "y_pred_base = knn_base.predict(X_test)\n",
    "y_pred_base_s = knn_base_s.predict(X_test_s)\n",
    "\n",
    "# create confusion matrices\n",
    "cm_base = confusion_matrix(y_test, y_pred_base)\n",
    "cm_base_s = confusion_matrix(y_test_s, y_pred_base_s)\n",
    "\n",
    "# show results for manual calculation\n",
    "print('Random Overall Accuracy: ', str(round(float(knn_base.score(X_test,y_test)) * 100)),'%')\n",
    "plot_confusion_matrix(knn_base, X_test, y_test)\n",
    "plt.show()\n",
    "print('Stratified Overall Accuracy: ', str(round(float(knn_base_s.score(X_test_s,y_test_s)) * 100)), '%')\n",
    "plot_confusion_matrix(knn_base_s, X_test_s, y_test_s)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The stock model on random sampling generalized better to the unseen data. Check the average cross_val_score for folds: 3,5,10 for each split."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Random Mean Score</th>\n",
       "      <th>Stratified Mean Score</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Folds</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.780488</td>\n",
       "      <td>0.723577</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.776571</td>\n",
       "      <td>0.760245</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.788833</td>\n",
       "      <td>0.781167</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       Random Mean Score  Stratified Mean Score\n",
       "Folds                                          \n",
       "3               0.780488               0.723577\n",
       "5               0.776571               0.760245\n",
       "10              0.788833               0.781167"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# empty storage lists\n",
    "random_cv_means = []\n",
    "stratified_cv_means = []\n",
    "\n",
    "# loop to get average values\n",
    "for fold_num in [3,5,10]:\n",
    "    #print(fold_num, ' - cv:\\n')\n",
    "    res = cross_val_score(knn_base, X_train, y_train, cv = fold_num)\n",
    "    res_s = cross_val_score(knn_base_s, X_train_s, y_train_s, cv = fold_num)\n",
    "    #print('random_scores: ', res,'\\n')\n",
    "    #print('stratified_scores: ', res_s, '\\n')\n",
    "    random_cv_means.append(np.mean(res))\n",
    "    stratified_cv_means.append(np.mean(res_s))\n",
    "\n",
    "# results\n",
    "avg_cvs = pd.DataFrame()\n",
    "avg_cvs['Folds'] = pd.Series([3,5,10])\n",
    "avg_cvs['Random Mean Score'] = pd.Series(random_cv_means)\n",
    "avg_cvs['Stratified Mean Score'] = pd.Series(stratified_cv_means)\n",
    "display(avg_cvs.set_index('Folds'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Check Precision and Recall via generating a classification report."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Random Sampling:\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "['              precision    recall  f1-score   support',\n",
       " '',\n",
       " '          ES       1.00      0.83      0.91         6',\n",
       " '          FR       0.57      0.80      0.67         5',\n",
       " '          GE       0.71      0.77      0.74        13',\n",
       " '          IT       0.50      0.33      0.40         6',\n",
       " '          UK       0.70      0.88      0.78         8',\n",
       " '          US       0.88      0.84      0.86        45',\n",
       " '',\n",
       " '    accuracy                           0.80        83',\n",
       " '   macro avg       0.73      0.74      0.73        83',\n",
       " 'weighted avg       0.80      0.80      0.79        83',\n",
       " '']"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# view the classifcation reports\n",
    "print(\"Random Sampling:\\n\")\n",
    "classification_report(y_test, y_pred_base).split('\\n')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Stratified Sampling:\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "['              precision    recall  f1-score   support',\n",
       " '',\n",
       " '          ES       0.86      0.86      0.86         7',\n",
       " '          FR       0.83      0.71      0.77         7',\n",
       " '          GE       0.75      0.75      0.75         8',\n",
       " '          IT       1.00      0.50      0.67         8',\n",
       " '          UK       0.53      0.91      0.67        11',\n",
       " '          US       0.85      0.79      0.81        42',\n",
       " '',\n",
       " '    accuracy                           0.77        83',\n",
       " '   macro avg       0.80      0.75      0.75        83',\n",
       " 'weighted avg       0.81      0.77      0.77        83',\n",
       " '']"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(\"Stratified Sampling:\\n\")\n",
    "classification_report(y_test_s, y_pred_base_s).split('\\n')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### KNeighborsClassifier() with tuned hyperparameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=5, estimator=Pipeline(steps=[('knn', KNeighborsClassifier())]),\n",
       "             param_grid={'knn__n_neighbors': range(1, 20),\n",
       "                         'knn__p': range(1, 3),\n",
       "                         'knn__weights': ['uniform', 'distance']})"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# set up the better format with the pipeline and steps\n",
    "# scaling is not necessary as they are all MFCCs\n",
    "steps = [('knn', KNeighborsClassifier())]\n",
    "pipeline = Pipeline(steps)\n",
    "\n",
    "# set up parameters and values to search over\n",
    "parameters = {'knn__n_neighbors': range(1,20), # set range through p (features)\n",
    "              'knn__p': range(1,3), # set range over manhattan and euclidean distances\n",
    "              'knn__weights': ['uniform', 'distance']} # use equal voting and weighting voting via proximity\n",
    "\n",
    "# instantiate the gridsearch cv object over the pipeline and parameters\n",
    "# this searches for the best parameters for the model and uses those for scoring\n",
    "knn_cv_object = GridSearchCV(pipeline, parameters, cv = 5) # use 5-fold cv as this works best in tuned model\n",
    "knn_cv_object_s = GridSearchCV(pipeline, parameters, cv = 5)\n",
    "\n",
    "# fit the object - random sampling\n",
    "knn_cv_object.fit(X_train, y_train)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=5, estimator=Pipeline(steps=[('knn', KNeighborsClassifier())]),\n",
       "             param_grid={'knn__n_neighbors': range(1, 20),\n",
       "                         'knn__p': range(1, 3),\n",
       "                         'knn__weights': ['uniform', 'distance']})"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# fit the object - stratified sampling\n",
    "knn_cv_object_s.fit(X_train_s, y_train_s)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Check the overall classification accuracy using the score method. It appears that the random sample produces much better accuracy than the stratified sample for the tuned model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The overall accuracy is:  83.0 %\n",
      "The overall accuracy is:  81.0 %\n"
     ]
    }
   ],
   "source": [
    "# random\n",
    "print('The overall accuracy is: ', round(knn_cv_object.score(X_test, y_test),2) * 100, '%')\n",
    "# stratified\n",
    "print('The overall accuracy is: ', round(knn_cv_object_s.score(X_test_s, y_test_s),2) * 100, '%')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The best parameters are:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Parameters - Random:  {'knn__n_neighbors': 6, 'knn__p': 2, 'knn__weights': 'distance'} \n",
      "\n",
      "Best Parameters - Stratified:  {'knn__n_neighbors': 3, 'knn__p': 2, 'knn__weights': 'distance'} \n",
      "\n"
     ]
    }
   ],
   "source": [
    "# print best params\n",
    "print('Best Parameters - Random: ', knn_cv_object.best_params_, '\\n')\n",
    "print('Best Parameters - Stratified: ', knn_cv_object_s.best_params_, '\\n')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Note:\n",
    "**Setting the cv value for the grid search object above to 3,5,10 yields the following values for score accuracy:**\n",
    "\n",
    "Random: [81%, 83%, 83%]\n",
    "\n",
    "Stratified: [81%, 81%, 73%]\n",
    "\n",
    "**This agrees with the original assumption I made that there wouldn't be enough observations in the training set for 10-fold CV to do well. This brings up an interesting point though. If the random state chosen pulled enough observations from the minority groups then the cross-validation would achieve good results, however if it failed to pull enough, then the generalization to unseen data might score really low. Because of this, the stratified sample should be used and the number of folds should be chosen to maximize the accuracy for the stratified sampling.**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Random Overall Accuracy:  83 %\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Stratified Overall Accuracy:  81 %\n"
     ]
    },
    {
     "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": [
    "# make the predictions and check the score method manually\n",
    "y_pred_tuned = knn_cv_object.predict(X_test)\n",
    "y_pred_tuned_s = knn_cv_object_s.predict(X_test_s)\n",
    "\n",
    "# create confusion matrices\n",
    "cm_tuned = confusion_matrix(y_test, y_pred_tuned)\n",
    "cm_tuned_s = confusion_matrix(y_test_s, y_pred_tuned_s)\n",
    "\n",
    "# show results for manual calculation\n",
    "print('Random Overall Accuracy: ', str(round(float(knn_cv_object.score(X_test,y_test)) * 100)),'%')\n",
    "plot_confusion_matrix(knn_cv_object,X_test,y_test)\n",
    "plt.show()\n",
    "print('Stratified Overall Accuracy: ', str(round(float(knn_cv_object_s.score(X_test_s,y_test_s)) * 100)),'%')\n",
    "plot_confusion_matrix(knn_cv_object_s,X_test_s,y_test_s)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "By tuning the hyperparameters of the model we achieved a 4% increase in accuracy for the proper stratified sampling."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Stratified Sampling:\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "['              precision    recall  f1-score   support',\n",
       " '',\n",
       " '          ES       0.75      0.86      0.80         7',\n",
       " '          FR       0.83      0.71      0.77         7',\n",
       " '          GE       0.86      0.75      0.80         8',\n",
       " '          IT       0.86      0.75      0.80         8',\n",
       " '          UK       0.69      0.82      0.75        11',\n",
       " '          US       0.83      0.83      0.83        42',\n",
       " '',\n",
       " '    accuracy                           0.81        83',\n",
       " '   macro avg       0.80      0.79      0.79        83',\n",
       " 'weighted avg       0.81      0.81      0.81        83',\n",
       " '']"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(\"Stratified Sampling:\\n\")\n",
    "classification_report(y_test_s, y_pred_tuned_s).split('\\n')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Visualize the decision boundaries"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Saving the following tuned model to the working directory..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['KNeighborsClassifier_tuned_model.sav']"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import joblib\n",
    "\n",
    "joblib.dump(knn_cv_object_s, 'KNeighborsClassifier_tuned_model.sav')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load and view saved model..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=5, estimator=Pipeline(steps=[('knn', KNeighborsClassifier())]),\n",
       "             param_grid={'knn__n_neighbors': range(1, 20),\n",
       "                         'knn__p': range(1, 3),\n",
       "                         'knn__weights': ['uniform', 'distance']})"
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_model = joblib.load('KNeighborsClassifier_tuned_model.sav')\n",
    "best_model"
   ]
  }
 ],
 "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": "3.9.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}