Unbalanced Data.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.tree import DecisionTreeClassifier, plot_tree"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_precisions_recalls(actual, preds):\n",
    "    plt.figure(figsize=(16,4))\n",
    "    \n",
    "    plt.subplot(1,2,1)\n",
    "    precision_0 = np.sum((actual == 0) & (preds == 0)) / np.sum(preds == 0)\n",
    "    precision_1 = np.sum((actual == 1) & (preds == 1)) / np.sum(preds == 1)\n",
    "    \n",
    "    plt.bar([0,1], [precision_0, precision_1])\n",
    "    plt.xticks([0,1], ['Class 0', 'Class 1'], fontsize=20)\n",
    "    plt.yticks(np.arange(0,1.1,0.1), fontsize=14)\n",
    "    plt.ylabel('Precision', fontsize=20)\n",
    "    plt.title(f'Precision Class 0: {round(precision_0,2)}\\nPrecision Class 1: {round(precision_1,2)}', fontsize=20)\n",
    "    \n",
    "    plt.subplot(1,2,2)\n",
    "    recall_0 = np.sum((actual == 0) & (preds == 0)) / np.sum(actual == 0)\n",
    "    recall_1 = np.sum((actual == 1) & (preds == 1)) / np.sum(actual == 1)\n",
    "    \n",
    "    plt.bar([0,1], [recall_0, recall_1])\n",
    "    plt.xticks([0,1], ['Class 0', 'Class 1'], fontsize=20)\n",
    "    plt.yticks(np.arange(0,1.1,0.1), fontsize=14)\n",
    "    plt.ylabel('Recall', fontsize=20)\n",
    "    plt.title(f'Recall Class 0: {round(recall_0,2)}\\nRecall Class 1: {round(recall_1,2)}', fontsize=20)\n",
    "    \n",
    "    plt.tight_layout()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Generate/Read Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "mode = 'read'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(314159)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "def generate_data(n, k, weights, cutoff):\n",
    "    X = np.random.random((n, k))\n",
    "    labels = (X@weights < cutoff*np.mean(X@weights)).astype(int)\n",
    "    return X,labels.flatten()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "def read_data():\n",
    "    df = pd.read_csv('diabetes.csv')\n",
    "    df = df.sample(frac=1).reset_index(drop=True)\n",
    "    df_train = df.iloc[:len(df)//2]\n",
    "    df_test = df.iloc[len(df)//2:]\n",
    "    X_train = df_train[[c for c in df.columns if c != 'Outcome']].values\n",
    "    labels_train = df_train.Outcome.values\n",
    "    X_test = df_test[[c for c in df.columns if c != 'Outcome']].values\n",
    "    labels_test = df_test.Outcome.values\n",
    "    return X_train, labels_train, X_test, labels_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "n_train = 10000\n",
    "n_test = 10000\n",
    "k = 10\n",
    "cutoff = 0.5\n",
    "weights = np.random.random((k,1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "if mode == 'generate':\n",
    "    X_train, labels_train = generate_data(n_train, k, weights, cutoff)\n",
    "    X_test, labels_test = generate_data(n_test, k, weights, cutoff)\n",
    "elif mode == 'read':\n",
    "    X_train, labels_train, X_test, labels_test = read_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fraction of positive labels: 36.458%\n"
     ]
    }
   ],
   "source": [
    "print('Fraction of positive labels:', str(round(100*np.mean(labels_train),3)) + '%')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Train Decision Tree"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf = DecisionTreeClassifier()\n",
    "clf.fit(X_train, labels_train)\n",
    "preds = clf.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "get_precisions_recalls(labels_test, preds)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Solution 1: Upweight the Minority Class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "weight_minority_class = np.sum(labels_train == 0) / np.sum(labels_train == 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.7428571428571429"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "weight_minority_class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf = DecisionTreeClassifier(class_weight={0:1, 1:weight_minority_class})\n",
    "clf.fit(X_train, labels_train)\n",
    "preds = clf.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABHgAAAEYCAYAAAAnPkG+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzdeZhcVbX+8e/LPInKEAhjNBcExatAQAaZZR6vPxUHkIgIMiooiooYVMABkCAggkpEQbmCilxEkClKGBPEAQFFSEDGRAYhEMb1+2PtIpXTVV3V1d1VaXw/z3OeSp+zz9m7qhrO6nX2oIjAzMzMzMzMzMxGrgV63QAzMzMzMzMzMxscJ3jMzMzMzMzMzEY4J3jMzMzMzMzMzEY4J3jMzMzMzMzMzEY4J3jMzMzMzMzMzEY4J3jMzMzMzMzMzEY4J3jMXmUkbSkpJE3o8Pzx5fzxQ9uy7pM0obyXLXvdFjMzM5s/NYt9JE2XNL0L9Yeka4e7HjN79XOCx6xN5eZbv70kaZakqyV9qNft+08gaQFJ75F0kaT7Jc2RNFvSHZLOkrRpr9s4VCTtIulaSU9KelrSTZL2GYZ69pF0c6njyVLnLgO8xvQG/31Uty82OG8LSf8n6V+SnpP0D0knSXrd0L1DMzMbbk1ipMfKPWW8JPW6jd0kaVVJX5M0TdLjkl6Q9KikKyV9QtJre93GoSBpGUmnlDjgOUkPSvqBpFWGuJ5VynUfLPVML/W+fgDXGK/WscpLDc5bVNLBJVaaVeKlOySdKmn1oXyfZkNhoV43wGwEOra8Lgy8CdgD2ErS+hFxRO+a9YqbgbWBWR2e/wvgRuChIWvREJC0InAhsCnwFPBb4B+AgDWADwAfk3RYRHy7Zw0dApIOAb4N/Av4MfA88B5gkqS3RsSnh6ieE4FPAf8EzgYWAd4PXCLp0Ig4rc1LnQI0SsoI+Bz538pllbo/BnwXeBH4OXA/sB5wBLCLpE0jotPfYTMz6436GOm/gP8BtgDGAYf0qlHdJGk/4DRgUeCPwE+Ax4FlgXeS98wvAsv1qo1DQdKywPXAmsDVwE+BtYCPADtL2jgi7hmCesaWekYBFwN3AhsCnwB2KPHCv9q41G3M/f2s2gzYmr6xykLAVWTseSf5XT4HbAAcCnxY0iYR8deBvi+zYRMR3rx5a2MDIv+T6bN/G+Dlso3pdTtfjRuwBHljDvLm+voGZZYGvgx8oW7fhHLOlr1+DwN4r2OAOWRyZ0zd/tcDd5f3s/EQ1LNJudbd9Z9nqf9fpQ1jBlnH9qWOWyv7VwSeBV4ANqwcO7KcM6nX34U3b968eWtv6ydG2hR4qcRIb+h1O/tp//jyHsZX9k8Hpg/gOh8s13kM2LlJmU2B2xp8ftf2+nMY4Gf23dLukyv7Dyv7fzNE9VxerndoZf/JZf+ZQ1DHDeVau1X2v7fsvxJYoHLs2HLsB73+Lrx5q988RMtskCLiKjKrLzKjP8/cL5I+WIbXPF0/jlvSEpI+J+m2MszoaUk3SPpAs7okbSfpktLN97kyTOliSe+qK9NwDh5JbyzDmO6W9GzpOv1nSWeWpzC1ck3n4JG0vnJ4VK3+GZLOkDS6QdlJ5TpjJB1Q6poj6ZHSjoF0Tz4ceBswBfhQRDxeLRAR/46IY4ATW11M0h6Sfizpb3Wf/TRJh0nq8/9FSStIOlHSXaX8E+XfkyS9sa6clEOerpc0s7zf+yVdLmnPNt/rvuRTv9MiYnrd+3scOL78+PE2r9Wf2jWOq/88S52nlzZ8ZJB17F9ev1vZvxOwGPDLiLi5cuwkYCbwQUnLDLJ+MzProYiYwtwYaf3qcUnvkHShpIclPV/umd+VtFKj6ymHBB0n6S+SnlEOLf6jcjjUknXl1pc0sRx7rNyP/64cBtz2sJ6BkPQasvctwPsj4tJG5cpn8o42rreSpGMkTan7fB6UdL6ktZucs5ukqyQ9pLlDpiZLOqhSrq2YsJ+2LQnsDcwGvlQ5fBqZGNu+PkbqRDl/u3K90yuHv1Tq37v+u++gjnWAjYAHgOp3Vmv/pRHxcuXYxeV1+U7rNhsOTvCYDY3a2PKo7P8U8APgPvKGdxmAco6R68g/2F8qZX5I3iTOl/TVPhVIx5JPMbYsryeR3UbXBvbqt3GZgLmF/IP9duBU4EfAveQNuk+CpsE1diG7yO5KPsk4GbgLOBCYKmlMk1O/UbY/kjfnB4CPkUPB2lVLFHylwQ12HhHxXBvX+xo5HOgmMhj7EbAUMJH8Hl4haQkysfQpYAbwHeD7wJ+B3YE31xU/DphE9lD5X/IzuhJYmXwK1I6ty+tvGhy7rFKm1sZaUu/aNuvoqJ6BkLQC+bvyNHB+5fCK5bVP1+3y/U4nu/dv3mn9ZmY236jFSC/Ms1P6CHl/3RG4hhy6NBXYj4wrVquUfwNwK/B5spfpd8j46Z/kg6D6P7Q/Rg45vgs4BziTHHp+BDClJGOG2nuAZYAbI+KK/gq2GatsDhwFPAFcBHyLHEL/HuAWSW+rLyxpfzLp8GbgEjJO/DWwOHUPbIYiJgQ2LtedEhFPVd7by0Dt/W9VaeO1GtjiF7U45Ipq/FfqnUL28t6ozes1ckB5/X5EVOfgub287tjgAWBtvsIrB1G32ZDzHDxmg6TsPfMmMrlzS+Xw1uRwmj9U9p8CrAt8NiK+UXetxYBfAp+XdGFE3Fb2bwccQ958N4uIByptaDWZXS3o+GRETKycuyTZdbq/97gUmbhYiBzu9Pu6Y58lEyZnkU9ZqjYC3hoR95XyC5FjtbeStGGDHhzVulcFViPnapncX9kB2Dki/lGpZwEyCPywpNMi4qZyaBtgLHBKRBxeOWcRsqdLzQFkAmudiHimUrbdsfZvKq9/qx6IiIckzQZWkbREtY52le98ZeDpiGg019Lfy+uanVy/2JdM0kyqBn/MnR/qDQ3atgA5TAxyLL+ZmY1QkjYn72vPk3ME1vavSfbunA5sUR/XSNqanGdvIjmHT82PgdWBz0fECZV6liMfKNScABxc/YNd0keB7wEHAV8f5Nuremd5vWqIrnc1sEL1HloSO1PI2GvHukMHkJ/z2yLi0co59THIoGLCommsUgxFHNFuPduVegb8uUtanHxI+jL5e1F1KTlP4LuBP0u6kvyM1ye/72+TD3DN5htO8JgNkOYOfaqfZFnAtyJiRqX4WdXkTun6uhcwtT65AxARc0rCZHtyHPdt5dCh5fVT1eROOe+fbTb/2Qbnzm7jvN3JyQF/Up/cKU4ih/tsK2m1WiKnzpfr90XEi5LOISe025C6gK+J2pOkf0XEnDba2lI1uVP2vSxpIvBh8vO/qVKk0Wf3PHmjr/cC2SurWrbdCYNrQ9eebHL8SWDJUq6W4KlNrN1uwqedOqDxxMktSRL5BBYy8Vd1OZmw20PSuIiYWnfsk8x9Cjss3ejNzGx4VGKk2iTLAj5deaBwYCnziWpcExFXS/oVsKuk10TEU5LWJ+eOu40GiZnqPbZBPFbzA7J37faNrjNItXil3ZisX9UkTd3+P0q6GthO0sIRUd8z6kUqPaXKOY1ikE5jQug8jvgw2eOmGisOdT3tel8599KIuL96MCJC0nvIh6xfZN5e21cB5zfo9WPWU07wmA1cbaxxkN1mf0926/xxg7KNkhcbAAsCfebJKRYur/Xjqzcq9TUaTtOOX5HDwU6XtD35B/YU4K8RUR1W1sh65fXq6oGSsPkd2etiXfretKdWzyFXTIL2/oBvNvytYyXJdiQ5F8wbyYRJvZXr/j2Z7JVzlKT1yO7OU8gJEqs39fPIZNztkn5Wzr0hIpoFJh01v7y+8nmUnjx3DmEdr1y6w/PeRX6ut1aSN3nRiBmSjiF/J6dI+jkZEL+9nPsn4L9pkCgzM7P5WnU+lgA+GhHnVPZvXF63kLRBg+uMImOlNYFpzB2Cc3mrodoAkhYme7S8n/yj/LXMOzXFyo3OG6ThiFd2Jh+ijSNX3ar+7bYcc1c9PY986Ha7pAvIGGRKRMysnDPYmLCtppfXea7X4CHgsNQzAM3mCsyLZ8/6c8meUgeTQ+CeISfKPhX4naT3RsTFjc436wUneMwGKCLUutQrHm6wrzZ53QZla2apun+/Dng8Ivo8bWlH+YN6Q3JVqR3IrqYA90s6MSJObXGJ2hOUZkun1/Y3eoLyRIN9L5bXBVvUC/BgeV1O0mKD7cVT5j+6hRwedDN5436stOl15LKbrwy7ioh/S9qIXC1hN/KpH8AsSWcAX617enY4uXT7vuS4+aOAFyX9mux9dXcbTXySDNheS65mVbV0ef13W2+4eR0w93utavXErJVawNSo9w4AEXGCpL+SPXZ2Ipdov51c7v5tZIKn4dNLMzObP9VipDLUZ2NyzrozJc2IiPqHRLVY6MgWl6zFQrX4ok8v5iYuIHsP3UP+Uf4wubw15H1n0SbnDUYtXmk1bL4tkg4jh6k9Tg5Zu49MLgTZe/xtzBuvnCxpFjn87DDyfYakycCRtQcuQxATQus4YulKuU4NWz2S3kz2Cvsn+fCukaPIORQ/ERH1SaDLSs+e28jvyAkem284wWM2vBo9UajdhL4VEUe0eZ0ngGUlLT6IJM8dwJ5lDpy3kT0lDgUmSpodEd/v5/Ram1dscnx0pdyQiYj7Jd1HzsOzOXMn7uvUfmRy59iImFB/QNLGZIKn2oZ/Ah8tQ4/eTM6tdDDZZXcBstsupUfPRPIzHUWOz34/GRy8RdJb2phY8S4ywbMmuWxnfftGk72N/tnp/DulnbMlPQCsLGl0g3l41iivzca8N1Xe9+40nly52o6LaRAUSTqw/LM6p5WZmY0AZajPlZJ2JSdG/qGkN9Xdu175wz0i2nlgUXtY1LLnjaRxZHLnSmCn+iFMZZ63z7T5NgbqOvIBzzaUuKBTJVY7lkxMrVe9T5d4pY+IOBc4tzzM2oT8HPYFLpe0dm3Y1yBjQshYBZrPsdNxHNHFevqbXLmmNpHyNdUDZajcY8DqkpaNiEYP5cy6zqtomXXfzeRkbpsN4JwbyW6oOwy28oh4MSKmRcTXyd4SkE+C+lObR2jL6oESHNQmFrx1sO1rotYT5OgGqxhU29Pqqdx/ldeLGhzbor8TI90eEd8Gti27G352EfFoRPw8It5HDm0bC6zTom0wdxhco+96x0qZwRiuej5CDjP8SYPJlVuStBb5+3QvlQSXmZmNLBHxJ+BssldL/UIFN5bXdmOhWvntW8UBzL3P/6oyPw3k3H+Lt1nnQF1I9gjeuCzA0VQbscpyZK+l6xskd5Zi7tD5hiLiiYj4dUR8jFwkYxkafNYdxoSQ38ezwKbVFcnK91NbdKNPYmSAaudvV/3eS72blnbcWD2xP2Xo1d5kPN5fMqv2PfVZCr18h7UeRNX5GM16xgkesy4rT0/OA8ZJ+mJJkMxD0ljlcqA13y6vJ0nq8/Sq0b7K8Q2Vy1ZX1fa16g3ySzJo+UAZrlTvk+R8K1cOw9jqmm+Ry6xvxtwnU/OQtFSZ1+XTLa41vbxuWTl/XeBzDa67jhovAT/PZydpUUnblF4+9ecvTAZWr5Rt4RyyG/kh9fVKej25NCzkcq/1dSwhaS1VlpRtoXaNL5Rr1641huyd9FxpS309o0s9DbtKVyZXbjieva7s0g32jSJ7/SxArjDXzkoeZmY2f/squaz5p+vuN6eRkwF/S7mi1jwkLSLplYREREwDrifnavtsg/LLlj/aofl9fhRw+qDeST/KQ43Dyo8XlPlt+ihxVKsHGI+SMcP6JaFTO3dhsqdwn5U5Je3QKKYk5zOCufHKYGNCIuJpcmn1JcmhXvUOIedlvDwi7qm0cbUSRyzRqo5Szz/InttjyNik3rGl/nPrJ4eWtHCpY2w/l34vOQ/krxtNrlyntrDI5xsk5SaQo2Fu6eSBltlw8RAts944hOxW+mVgb0nXAY8AK5GTK29APkm5FyAirpD0FbLL7x2SfklOVLwC2dvhRmB8P/V9EDi4jMO+mxzPPRbYlfxD/pT+GhsRT0vaF/gZMFk5gfB95DKR25FdiA/o5xKDEhHPSNqBfDr2IXJljd+W97IA+bRuG/JJyiEtLncuOeb/FElbkUtsrkF2w/05sGel/LuAkyVdT05k/Cj5JHJ38snPN0u5xcnu4NMl3QTMABYje/qsTT5JvKON93qvpCPJyfumKidKfJ5c1nQV4KSIqAaGG5JPuSbToJdVk3qul3QycATwJ0kXkvPg7EkmpA6NiOmV004A9iF76UxqcNmtye/i1hKM9+eY8p3eAMws7203cpz9MRHxs3beh5mZzd8i4gFJ3yWHQH8G+FxE3Fniih+QkwL/hhxmszA5JHsz8t6wVt2l9gKuBY6X9P/Kv0Xew7crZaeTw3unAO8u9+7ryHhpR3LIz4MMk4g4T7n09mnAbyTdRiamHifnHdqYHBLV78qaZWXPU8k5YP4s6WLyHr0VeY++pvy73k+BOSWmnE5+NpuRMeU0MkaBQcaEdT5PxhxHSHo7c1f03J2MlaoJGcgYbIvS9mvbrOcg8jM8VdI2wB3AO8o1/gZ8oVJ+5VJmBpkYaqTlXIHFceTnsg1wZ/k9fZbsObRh+Xefof1mPRUR3rx5a2Mj59OJNstOKOW37KfMImQy4npyLPpzZNLkKrJXzLINztmJXEnrsVL+fuAXwNZ1ZbYsdU+o2/cO4DtkL5jHyBvS3WQPjXUqdYwv549vUP8Gpb6ZZNLhvnLdlRqUnVSuM6bBsT5tbPNzXYB86lJbdWkO+aTpTuB7wCbtfA/kPDq/IgOQ2WTgsx8ZCAQwqa7s2uSSqlPL+36ODJwurK+PDEo/A1xWPpc5pfyN5AoYiwzwve5KJmyeKm28BdinSdna53ltB7/X+5Rrzy51TQZ2aVK29p32+d0oxy8oxw9oo96dySFgj5bfpUfK97rZcPz3682bN2/ehm+jRYxEJlhml22Fuv1vLfeWGeX++hjwF7IX6NYNrrMsubz5XeU++wQ50e1xwBJ15ZYBzij36znkAgjHk0t0TwemV647vtH9rVHZNj+PVUs7by1tfKHEBNeQMd7SDT6/ayv7FiIfwvyVjNseJnvNrE6DGKvEGr8gJ5Z+pnyWfyixyWvqyrUdE7bxPpchexTNKPfyh8ik3SpNyl9Li/i4n8/znHL950t9E4FlGpQdU+po+L2RcV2QMfSCbdS9PHAimTSaU1f/OcBavf5vz5u36qaIIVvJz8zMzMzMzMzMesBz8JiZmZmZmZmZjXBO8JiZmZmZmZmZjXBO8JiZmZmZmZmZjXBO8JiZmZmZmZmZjXBO8JiZmZmZmZmZjXBO8JiZmZmZmZmZjXBO8Ji9ykjaUlJImtDh+ePL+eOHtmXdJ2lCeS9b9rotZmZmNn9qFvtImi5pehfqD0nXDnc9Zvbq5wSPWZvKzbd+e0nSLElXS/pQr9v3n0DSApLeI+kiSfdLmiNptqQ7JJ0ladNet3GwJL1O0pGSzpP0V0kvlt+3dw1TfftIulnS05KelHStpF0GeI3xDf77qN8+Xik/pkX52rbZ0L5bMzMbDk1ipMfKPWW8JPW6jd0kaVVJX5M0TdLjkl6Q9KikKyV9QtJre93GwZK0raSTJF1VvuuQdN0w1bWKpB9IelDScyXxdoqk1w/wOtP7iTkebnLOopIOLrHSrBIv3SHpVEmrD807NBs6C/W6AWYj0LHldWHgTcAewFaS1o+II3rXrFfcDKwNzOrw/F8ANwIPDVmLhoCkFYELgU2Bp4DfAv8ABKwBfAD4mKTDIuLbPWvo4I0BvlH+/U/ye1xhOCqSdCLwqVLP2cAiwPuBSyQdGhGnDfCSFwO3Ndg/tfLzE8z976hqVWBf4F/k77KZmY0c9THSfwH/A2wBjAMO6VWjuknSfsBpwKLAH4GfAI8DywLvBE4Bvggs16s2DpGDgd2BOcDdwICSLe2SNBa4HhhFxhl3AhsCnwB2kLRpRPxrAJd8kvwOqp5uUPdCwFVk7Hkn+V0+B2wAHAp8WNImEfHXAdRvNqyc4DEboIiYUP+zpG3IZMMnJZ0aEdN70a6aiHiGvAl1ev6T5M1vviFpCeA3wNuAnwIHRcTjlTJLA58Glu5+C4fUDOBdwB8i4jFJk4B9hroSSZuQyZ1/ABvUPk9J3wSmASdK+r8B/j7/MiImtSoUEU8AE5q064Tyz3Mj4rkB1G1mZj3WIEbaFPgdcJCkkyLi3p40rEskfZB8YPI48P8i4tIGZTYFTu9224bB14EvkDHnqsBwfbdnkMmdeR7gSToZOBw4Dvh4k3MbeaL6e9qP/yGTO1cB20XEy3X1HwscQ8ae+w6gfrNh5SFaZoMUEVeRNzeRGf155n6R9EFJN5UundNr50laQtLnJN1Whhk9LekGSR9oVpek7SRdUrr5PqccpnRx/fAdNZmDR9IblcOY7pb0bOlO+2dJZ0patq5c0zl4JK2vHB5Vq3+GpDMkjW5QdlK5zhhJB5S65kh6pLRjIN2TDyeTO1OAD1WTOwAR8e+IOAY4sdXFJO0h6ceS/lb32U+TdJikPv9flLSCpBMl3VXKP1H+PUnSG+vKSTnk6XpJM8v7vV/S5ZL2bOeNRsTjEXFVRDzWTvlBqAVDx9V/niWhczr55PEjw9yGeZQnZePLj2d1s24zMxt6ETGFuTHS+tXjkt4h6UJJD0t6vtwzvytppUbXk7SMpOMk/UXSM8qhxX9UDodasq7c+pImlmOPlfvx35VDioarp8lrgFoC4v2Nkjvwymfyjjaut5KkYyRNqft8HpR0vqS1m5yzm3LI1EMlTntQ0mRJB1XKtRUT9iciboiI2yPipXbKd6LEWNsB0+mbFPsSMBvYu/67H2K1GO/S+uROcXF5XX6Y6jbriHvwmA2N2tjyqOz/FLAtcAlwDfBayHlWgKuBdYFbgR+QCdftgfMlvSUijp6ngrlPCp4GfgncD6wEbALsBVzZtHGZgLmF7N3ya+AiYDHgDcDeZFfifru3Kudluai81wvJnibrAwcCuyu7yE5vcOo3yvu6BLgC2Ar4GNl1e+v+6qyzf3n9SoMb7Dza7PXxNeBl4CbgAfJ72RqYSCbp9q4VVPYemgKMJXtqXUJ+BquTXZMvBO4pxY8DPkc+xfpfsifU6HLN9wIXtNG2AVNOIn0NMDkitmzztNpn/5sGxy4ju49vTQZQ7Xq7pE+Sv1sPANdExD8HcP7uwIrA7yKi415oZmY2X6nFSC/Ms1P6CNnb5TngV2RcswawH7CrpI0i4r668m8g73Wrkz1Nv0PGTmuSD4LOJP/gh4wz/geYTMZHCwLrAUcAO0p6R0Q8NcTv8z3AMsCNEXFFfwXbjFU2B44i3/NFZPy3RqlntxJ3/bFWWNL+wHeBh8lYZRbZ8+W/yQc2Z5Ryg44JO6WcSHoLYKuIuLaNU2qxyhXV+C8inpI0hUwAbUT2smnHopL2AlYjf1/+RMYdjRJVt5fXHSVNrLShNl9h0/jbrBec4DEbJGXvmTeRyZ1bKoe3BjaOiD9U9p9CJnc+GxG1+VaQtBiZvPm8pAsj4rayfzsyuXMvsFlEPFBpwyotmlkLOj4ZERMr5y5JJjv6e49LAZPI/2dsGRG/rzv2WTJhchZ5k63aCHhrLUgrvTSuJuct2jAi+p1nRdKq5E34RTJQGwo7R8Q/KvUsAJxDjqc+LSJuKoe2IZM7p0TE4ZVzFiF7utQcQCY21ilD5erLzjdj7ct3vjLwdEQ0mmvp7+V1zQFe+hOVn1+S9D3y925OG+fXEnnfHWC9ZmY2H5K0ORkjPU/dvGqS1iT/Xz8d2KI+rpG0NflAZSKZpKn5MZnc+XxEnFC3v3aPrZ9D5QTg4Oof7ZI+CnwPOIgcYjSU3lle2000tHI1sEI1ESWp1qP5a8COdYcOID/nt0XEo5Vz6mOQQcWEXfam8vq3Jsf/Tsaea9L+574i8KPKvnslfSQiqnHmpcDPgXcDf5Z0JfkZr09+398mE2Jm8w0P0TIbIOXwqwmli/CFZA8IkQmAGZXiZ1WTO6Xr617A1PrkDkD5I/iz5XofrDt0aHn9VDW5U85rt5fEsw3OnR0RffZX7E5ODnhBfXKnOIkM0LaVtFqDc79c/wQuIl4kEymQk+S1Uhv+9a82kwQtVZM7Zd/LZDAJ2eOoqtFn93yDJ4AvAH2eAkVEp5Net6M2sfaH2yxfGx7XbK6l2v7XtXm9e8nf0TcBS5I9y95H/l4cQPZQ65ekMeTcQ/8inyaamdkIU4mRLiB7Nwj4dOWBwoHkRMyfqMY1EXE12aNn1zLsCUnrkz2Wb6NBYiYiZtXHCBExo0mPjB8A/6bxfX6wavHKQHquNhURjzbqZVR67dQelC1cOfwilZ5S5ZxGMUinMeFgfJiMV9pdRGGo45VzyAd3K5LxylvJROMY4LKSPHtFRASZEJtAxjiHkXPubEXOLXX+cA5RM+uEe/CYDVxtyEqQqwH9Hvh+RPy4QdlGN7ANyK7CfebJKWo36/rx1RuV+hoNp2nHr4DjgdMlbQ9cTj79+Wu5ebWyXnm9unogIl6U9Dvy5rgucF+lSHUFJchu2NDeigvNhr91rCTZjgR2IsdXV8dur1z378lkr5yjJK1HdmeeAtzW4KZ+HpnouF3Sz8q5N5SJq4fNYCfW7u/SbdY/mXl7Vz0D/EzSjeQKIh+Q9PX6ruQNfIx86PBDT65sZjZiVYf1BvDRiDinsn/j8rqFpA0aXGcUGSutSQ7H2qjsv7zVUG2Akvg4gFwZ8s1koqD+wfbKjc4bpOGIV3Ym58wbR666Vf3bbTnmrnp6HvnQ7faSXJsMTImImZVzBhsTdqz+gd8QGdBnHhHVFTz/Anxc0tPktAoTqOs1VnrWn0v2lDqYnHfnGXLi5VOB30l6b0RcjNl8wgkeswGKCLUu9YqHG+yrTV63QdmaWaru368DHu/0qUpEzJC0IXnj2oHsagpwv6QTI+LUFpeoPUFptnR6bX+jJyhPNNj3YnldsEW9AA+W1+UkLTbYXjxl/qNbyLHmN5M37sdKm15HDjN6ZdhVRPxb0kbk0q+7Mfep3yxJZwBfjYja07LDyVWp9iXHzR8FvCjp12Tvq7sH0/YhVEs4NZvoutUTs7ZExP3lvX+InGytT+sAACAASURBVEugYYKnDNurTejsyZXNzEaoWoxUhvpsDHwfOFPSjNIzp6YWCx3Z4pK1WKgWX/TpxdzEBeQf6veQf5Q/TM71A/BJ5h1ePVRq8UqrYfNtkXQY2bP4cXLI2n1kciGAPcjFJ+rjlZMlzSKHnx1Gvs+QNBk4MiKmlnKDjQm7qVW8snSlXKfOJBM8m1f2H0XOofiJiKgfPn6ZpPeQPcomMnfCZbOec4LHbHg1eqJQuwl9KyKOaPM6TwDLSlp8EEmeO4A9yx/TbyOHwxwKTJQ0OyK+38/ptTav2OT46Eq5IVOSBPeR8/BsTk7UPBj7kcmdY6Pvcq4b03cemdoQuI9KEvkkcGvySc4x5BPBL5ZyL5E3+omSRpHjs99PBgdvKZNn97x3SkTMlvQAsLKk0Q3m4VmjvDYb8z4QtSeH/a1wsSv5OzQ5Iu4agjrNzKyHImI2cKWkXcnFJH4o6U1189O98od7RPy7jUvWHha17HkjaRyZ3LkS2KnuIUxtvr3PtPk2Buo68gHPNpS4oFMlVjuWTEytV71Pl3ilj4g4Fzi3PMzahPwc9gUul7R2bW6eQcaE3VSLCZrNCThU8UptzqJqrFKbSPma6gkR8UdJjwGrS1o2IoZlYmqzgfIcPGbddzM5gd1mAzjnRrIb6g6DrTwiXoyIaRHxdaC2JPseLU6rzSO0ZfVACQ5qEwveOtj2NVHr1XG0GixjXmlPq6dy/1VeG83zskV/J0a6PSK+Ta6OBk0+uzJ2/ucR8T5yaNtYYJ0Wbeum2pPURr9TO1bKDEZtKdh7+ilTm1zZvXfMzF5FIuJP5EpZq5C9XGtuLK/txkK18tu3igOYe5//VX1yp9gQWLzNOgfqQrJH8MZlAY6m2ohVliN7LV3fILmzFHOHzjcUEU9ExK8j4mPkIhnL0OCz7jAm7KZaYmW76vde5mfalJxL6MbqiQNUS5hVY5Xa99RnKfTyHdZ6ED0/yPrNhowTPGZdVp6enAeMk/TFkiCZh6SxZTnQmm+X15Mk9Xl61Whf5fiGklZocKi275kGx+r9kgxaPlCGK9X7JDmPzZXDMLa65lvk8J7NmPtkah6SlpJ0DDn5XX+ml9ctK+evSy5xXr3uOmUC4Kp5PjtJi0rapvTyqT9/YTKweqXsUJO0hKS1mkxy3cyZ5fULkl6ZC6m814PJruzzzJkgaXSp57WV/X2CRqXPkUHTLJrMHyVpdXIFDE+ubGb26vRVYA7w6br7zWnkZMDfUq6oNQ9Ji9TfWyJiGnA98HZyMYpq+WXLfCnQ/D4/Cjh9UO+kH2VC5MPKjxeU+W36KHHUDS0u9ygZM6xfEjq1cxcmewr3WZlT0g6NYkpyPiOYG68MNibsmKTVShyxRDvly6IYV5DzPB5cOXws2ePm3NJjrFbHwqWOsZW63yJpmco1anFIbSWs6nyatYVFPt8gKTeBHA1zS6PJsM16xUO0zHrjELJb6ZeBvSVdBzxCrj60Njk3zwfI1YmIiCskfYXs8nuHpF+SExWvQPaeuREY3099HwQOLuOw7ybHc48lh8Y8Ry7b3lREPC1pX+BnwGTlBML3kctEbkd2IT5gYB9B+yLiGUk7kE/HPkSurPHb8l4WIJ/WbUM+STmkxeXOJcf8nyJpK3KJzTXIbrg/B/aslH8XcLKk68mJjB8ln0TuTvbE+mYptzjZHXy6pJuAGcBiZE+ftckniXe0834lncjc4K3WO+pISXuVf/8yIn5Zd8qG5FOuyTToZdVIRFwv6WTgCOBPyhXhFiHf/zLAoRExvXLaCcA+5Hw5k+r2/07S38i5jR4gx8pvSvZYegb4UD9d8PfDkyubmb1qRcQDkr5LDoH+DPC5iLizxBU/ICcF/g05zGZhckj2ZuQQ37XqLrUXcC1wvKT/V/4t8h6+XSk7nbwXTQHeXe7d15Hx0o7kkJ8HGSYRcZ6kxcmEwW8k3UYmph4n5x3amBwS1e/KmhHxsqRTyTlg/izpYvIevRV5j76m/LveT4E5JaacTn42m5Ex5TQyRoFBxoQ1kt5J3sNh7lxJa0iaVPc+xldOO5fsLb0V+f214yDyMzxV0jbAHWTv4K3I35kvVMqvXMrMIBNDNe8lF8y4hoyvnyLf985kvPZr4MTKtY4jP5dtgDvL7+mzZIyzYfl3n6H9Zj0VEd68eWtjI+fTiTbLTijlt+ynzCJkMuJ6ciz6c2TS5CqyV8yyDc7ZiewJ8Vgpfz/wC2DrujJblron1O17B/AdshfMY+QN6W6yh8Y6lTrGl/PHN6h/g1LfTLI76n3luis1KDupXGdMg2N92tjm57oAeYP+ObkM6RwygXAn8D1gk3a+B3IenV+RyZrZZOCzHxkIBDCpruzawMnkamAzy+c+nUw2bVJXbmEycL2sfC5zSvkbyRUwFhnA+5xe+31rsk2olK99ntd28Hu9DxkMzyaDncnALk3K1r7T8ZX93yznPVj5Tk4D3thP3QuSCaEA3tSL/669efPmzdvgN1rESGSCZXbZVqjb/9Zyb5lR7q+PkSsbfZe62Kau/LLkMul3lfvNE+REt8cBS9SVWwY4o9xP55ALIBwPLFH2Ta9cd3yT+1ufsm1+HquWdt5a2vhCiQmuIWO8pRt8ftdW9i1EPoT5Kxm3PQz8CFidBjFWiTV+QQ4zeqZ8ln8gY5PX1JVrOyZs8R5rn1nTrcE519IiPu7n8zyHXNTj+fL7MhFYpkHZMaWO6ne8BfATMj6p/05+Sy7friZ1L08mfu4ov0u1+s8B1ur1f3vevFU3RQzranhmZmZmZmZmZjbMPAePmZmZmZmZmdkI1/MEj6TNJf1K0gOSQtL4Ns55q6TJkp4t5x1TndjUzMzMrNcc55iZmVm39DzBQ07K9RdygqpnWxWWtDQ5VvIRcj6Qw8gJU48YxjaamZmZdcJxjpmZmXXFfDUHj6SngUMiYlI/ZQ4kJy1bISKeLfuOBg4EVon56Q2ZmZmZFY5zzMzMbDiNxGXSNwZ+Xwt6isuBr5Czpt9bX1jS/sD+AEsuueT6a61Vv9qimZmZvdpMmzZtVkQs3+t2dGhAcQ441jEzM/tP0yzWGYkJnhXJ5ZHrPVJ3bJ7AJyLOAs4CGDduXEydOnXYG2hmZma9I2lGr9swCAOKc8CxjpmZ2X+aZrHO/DAHTyeq3ZPVZL+ZmZnZSOM4x8zMzAZsJCZ4HiafYNUbVV4fwczMzGzkcpxjZmZmHRmJCZ4bgM0kLVa3b1vgQWB6T1pkZmZmNjQc55iZmVlHep7gkbSUpLdLentpz2rl59XK8RMkXVV3yvnAM8AkSetIejdwFHCyV5YwMzOz+YnjHDMzM+uWnid4gHHAH8q2OHBs+feXy/HRwNha4Yh4knyStRIwFTgdOAk4uXtNNjMzM2uL4xwzMzPrip6vohUR1zJ38sBGx8c32PdnYPPha5WZmZnZ4DnOMTMzs26ZH3rwmJmZmZmZmZnZIDjBY2ZmZmZmZmY2wjnBY2ZmZmZmZmY2wjnBY2ZmZmZmZmY2wjnBY2ZmZmZmZmY2wjnBY2ZmZmZmZmY2wjnBY2ZmZmZmZmY2wjnBY2ZmZmZmZmY2ws0XCR5JB0m6V9IcSdMkbdai/Psk3SbpGUkzJB3ZrbaamZmZDZRjHTMzMxtuPU/wSNoTmAgcD6wLXA9cJmm1JuV3BM4HzgLWAQ4CDpd0SHdabGZmZtY+xzpmZmbWDT1P8ABHAJMi4uyIuCMiDgUeAg5sUn5v4JKIOCMi7omIS4ETgM9KUpfabGZmZtYuxzpmZmY27Hqa4JG0CLA+cEXl0BXAJk1OWxSYU9n3LLAKsPqQNtDMzMxsEBzrmJmZWbf0ugfPcsCCwCOV/Y8AKzY553JgD0nbSVpA0prAp8qx0dXCkvaXNFXS1JkzZw5Vu83MzMza4VjHzMzMuqLXCZ6aqPysBvtqzgZOBS4GngduBH5ajr3U58IRZ0XEuIgYt/zyyw9Rc83MzMwGxLGOmZmZDateJ3hmkYFK9QnWKPo+6QIg0meBpchuyisCN5fD04enmWZmZmYdcaxjZmZmXdHTBE9EPA9MA7atHNqWXGGiv3NfiogHyjU+ANwQEY8OT0vNzMzMBs6xjpmZmXXLQr1uAHAy8CNJNwNTgI8DKwFnAkg6AdgwIrYpPy8HvBe4lpyE8CPl5y263nIzMzOz1hzrmJmZ2bDreYInIi6QtCxwNDlx4F+AnSJiRikyGhhbOe3DwDfJ8es3AFtGxM2YmZmZzWcc65iZmVk39DzBAxARZwBnNDk2vvLzLGDjLjTLzMzMbEg41jEzM7Ph1utJls3MzMzMzMzMbJCc4DEzMzMzMzMzG+Gc4DEzMzMzMzMzG+Gc4DEzMzMzMzMzG+Gc4DEzMzMzMzMzG+Gc4DEzMzMzMzMzG+Gc4DEzMzMzMzMzG+Gc4DEzMzMzMzMzG+Gc4DEzMzMzMzMzG+HmiwSPpIMk3StpjqRpkjZrUX57STdIekrSLEkXS1qzW+01MzMzGwjHOmZmZjbcep7gkbQnMBE4HlgXuB64TNJqTcq/AbgY+H0p/y5gceDXXWmwmZmZ2QA41jEzM7Nu6HmCBzgCmBQRZ0fEHRFxKPAQcGCT8usDCwOfi4i7I+I24ARgrKTlutNkMzMzs7Y51jEzM7Nh19MEj6RFyCDmisqhK4BNmpw2FXgB2E/SgpJeA+wD3BIRs4atsWZmZmYD5FjHzMzMuqXXPXiWAxYEHqnsfwRYsdEJETEd2BY4FngOeBJ4K7BLo/KS9pc0VdLUmTNnDlGzzczMzNriWMfMzMy6otcJnpqo/KwG+/KAtCLwfeBcYANgS+Ap4H8l9Xk/EXFWRIyLiHHLL7/8kDbazMzMrE2OdczMzGxYLdTj+mcBL9H3CdYo+j7pqjkYmB0Rn6ntkLQXcD/Z1fm6YWhnW8YcdWmvqjYbkaZ/bedeN8HMbLi9qmIdMzMzm3/1tAdPRDwPTCO7IdfbllxhopElyECpXu3n+aVHkpmZmZljHTMzM+uaXvfgATgZ+JGkm4EpwMeBlYAzASSdAGwYEduU8pcCh0v6EnA+8Bpy2dH7yQDKzMzMbH7yqoh13FPZbGDcU9nMuq3nCZ6IuEDSssDRwGjgL8BOETGjFBkNjK0rf7WkDwKfAY4EngVuBHaIiNldbbyZmZlZC451zMzMrBt6nuABiIgzgDOaHBvfYN9PgZ8Oc7PMzMzMhoRjHTMzMxtuHsdtZmZmZmZmZjbCOcFjZmZmZmZmZjbCOcFjZmZmZmZmZjbCOcFjZmZmZmZmZjbCOcFjZmZmZmZmZjbCOcFjZmZmZmZmZjbCOcFjZmZmZmZmZjbCOcFjZmZmZmZmZjbCOcFjZmZmZmZmZjbCzRcJHkkHSbpX0hxJ0yRt1k/ZCZKiyTaqm+02MzMza4djHTMzMxtuPU/wSNoTmAgcD6wLXA9cJmm1JqecCIyubJOBayPi0eFvsZmZmVn7HOuYmZlZN/Q8wQMcAUyKiLMj4o6IOBR4CDiwUeGIeDoiHq5twMLAZsDZ3WuymZmZWdsc65iZmdmw62mCR9IiwPrAFZVDVwCbtHmZjwJPABcNYdPMzMzMBs2xjpmZmXXLQp2eKGkcsCHwemDBBkUiIr7S4jLLlXMfqex/BHhXG21YANgXODcinmtSZn9gf4DVVmvWE9rMzMxsWDjWMTMzs64YcIJH0tLAz4GtAPVTNIBWCZ76svNU02BfIzsBqwLfa3rhiLOAswDGjRvXzjXNzMzMhppjHTMzMxtWnfTg+SawNfB74BzgfuDFDuufBbwErFjZP4q+T7oa+RhwfUTc3mH9ZmZmZsPJsY6ZmZl1RScJnt2BW4GtIuLlwVQeEc9LmgZsC/ys7tC2tBhnLmklYGdgv8G0wczMzGy4ONYxMzOzbulkkuXXAtcMNrlT52RgvKT9JK0taSKwEnAmgKQTJF3V4Lx9gdnA/w5RO8zMzMyGg2MdMzMzG3ad9OD5O7DCUDUgIi6QtCxwNDAa+AuwU0TMKEVGA2Prz5EkckWJ8yLimaFqi5mZmdlQc6xjZmZm3dBJgud04GuSVo6IB4aiERFxBnBGk2PjG+wL4A1DUbeZmZnZcHOsY2ZmZsOtkwTPZeQky1MkHQtMA55oVDAi7htE28zMzMzMzMzMrA2dJHimk8t6in6W7CxlOrm+mZmZmZmZmZkNQCcJmHPJ5I2ZmZmZmZmZmc0HBpzgaTRO3MzMzMzMzMzMeqeTZdLNzMzMzMzMzGw+Mqg5ciStAqwLvA54Erg1Iv45FA0zMzMzMzMzM7P2dJTgkbQacBawbYNjvwU+HhHTB9c0MzMzMzMzMzNrx4ATPJJWBKYAK5Mrav0OeAgYDbwT2A64TtK4iHh46JpqZmZmZmZmZmaNdNKD54tkcuezwMkR8VLtgKQFgcOBbwBHA4cMRSPNzMzMhoukezo8NSJi7JA2xszMzKxDnUyyvDNwRUR8sz65AxARL0XEicAVwC7tXlDSQZLulTRH0jRJm7UoL0mflHSnpOckPSTpax28FzMzM7MFAHWwtR1HOdYxMzOz4dZJD54VgfNalJkGbNnOxSTtCUwEDgKuK6+XSXpzRNzX5LSTyATSkcCfgdeSQ8TMzMzMBiQixgzn9R3rmJmZWTd0kuB5Eli9RZnVSrl2HAFMioizy8+HStoBOBD4XLWwpDcBhwL/HRF31B36Q5v1mZmZmXWTYx0zMzMbdp0M0boOeI+kTRodlPQO4L2lXL8kLQKsTw7pqncF0PD6wO7APcAOku6RNF3SDyWNalLH/pKmSpo6c+bMVk0yMzMzGzKOdczMzKxbOunBcxw5D89kST8FriFX0VqRHJb1AeBl4Pg2rrUcsCDwSGX/I8C7mpzzRrIH0fuB8UAAJwKXSNo4Il6uLxwRZ5FLujNu3Lhoo01mZmb2H0TShzs9NyLObVHEsY6ZmZl1xYATPBFxq6T3AJOADwEfrDss4DFg34iYNpDLVn5Wg301CwCLAntHxN8AJO0N3AVsANw0gHrNzMzMJtE87mimFqu0SvDUONYxMzOzYdVJDx4i4v8krU52IV6PnPjvSXJs+C8jYnabl5oFvET2/qk3ir5PumoeAl6sBTzF34EXybl/HPSYmZnZQHxkGK/tWMfMzMy6oqMED0BJ4pxftk6v8bykacC2wM/qDm0LXNTktCnAQpLGRsQ/yr43ku9lRqdtMTMzs/9MEfHDYby2Yx0zMzPrio4TPEPoZOBHkm4mA5qPAysBZwJIOgHYMCK2KeWvBG4FfiDpk2XfKeTTrKndbLiZmZlZGxzrmNmINuaoS3vdBLMRZ/rXdu56nS0TPHUTD/4iIp4ayESEbUw8SERcIGlZ4GhgNPAXYKeIqD2hGg2MrSv/sqRdgFOB3wHPAr8FjqhOOmhm1i0OfMwGphdBT6841jEzM7NuaKcHzyRyEsAbgadobyLCAU08GBFnAGc0OTa+wb6HyKXYzczMzIacpCWBg4DtgZXJSY+rIiLGNtjfqKBjHTMzMxtW7SR49iWTNQ+Vn4dzIkIzMzOznpL0OuA64M3Av4GlycUkFgEWL8UeBF7oSQPNzMzMGmiZ4ImISZWfh20iQjMzM7P5wNFkcuejZM/ll4BvAV8B3gGcBswme/eYmZmZzRcW6HUDzMzMzOYzuwG/i4hzIuKVYemRbgR2AtYCvtCrBpqZmZlVDTjBI+n1kt4sadHK/o9IuljS+ZLeMXRNNDMzM+uqVclVrGpepm4Onoh4FLgMeH+X22VmZmbWVCfLpB8P7AWMqu2QdCi5fKfKrj0kjYuIvw6+iWZmZmZd9Qw5LKvmSWDFSplHyMmXzczMzOYLnQzR2hS4KiKerdv3aeABYHPgfWXfEYNsm5mZmVkv3E/24qn5K7C5pAXr9r0TeLirrTIzMzPrRycJnpWBe2s/SHozGQR9OyKui4gLgUvIZI+ZmZnZSDMZ2EJSrWfyBcBY4FJJB0v6GbAR8OteNdDMzMysqpMhWosDc+p+3pRcRv3Kun3/AHYZRLvMzMzMeuWH5JLoq5C9ec4Etgb2ALYrZaaQq22ZmZmZzRc66cHzALlyRM32wL+BP9btez1QP4SrX5IOknSvpDmSpknarJ+yYyRFg22Hgb4RMzMzs6qIuDUiDoyI+8vPL0bEu4ENgA8AGwNbRMQT7V7TsY6ZmZkNt0568FwD7CPpELInz27ARRHxcl2Z/yKfeLUkaU9gInAQcF15vUzSmyPivn5O3YF5k0qPtf8WzMzMzAYmIqYB0wZ6nmMdMzMz64ZOevCcADxNBipnkUmeCbWDkkYBWwDXt3m9I4BJEXF2RNwREYcCDwEHtjjvXxHxcN32/ADfh5mZmVkfkhaXtJqkRZocX7QcX6zNSzrWMTMzs2E34ARPRNwLvAX4BHAYsE5E3FVXZHXgdGBSq2uVwGl94IrKoSuATVqc/nNJj0qaIuk9bTbfzMzMrJVjgLuApZocXxK4E/h8qws51jEzM7Nu6aQHD+Up0mllu69y7JaIODwibmnjUssBCwKPVPY/AqzY5JynyWXZ3wfsBFwFXCBpr0aFJe0vaaqkqTNnzmyjSWZmZvYfbkfgyohoOCSq7L+S9haUcKxjZmZmXdHJHDzDISo/q8G+LBgxCzipbtdUScsBnwF+3KD8WeRQMsaNG9fwmmZmZmZ1xpBJlf78DXjnAK7pWMfMzMyGVcsEj6RjyADk9Ih4rPzcjoiIr7QoMwt4ib5PsEbR90lXf24CPjKA8mZmZmbNLAy83KJMAO3MweNYx8zMzLqinR48E8gg5gJy9YYJbV47gH4TPBHxvKRpwLbAz+oObQtc1GY9AG8nJys0MzMzG6x7yAUj+rMlMKPVhRzrmJmZWbe0k+DZqrzeV/l5qJwM/EjSzcAU4OPASsCZAJJOADaMiG3Kz/sALwB/IJ+u7QocDHx2iNtlZmZm/5l+BRwl6TMR8Y3qQUlHAesBfY414VjHzMzMhl3LBE9ETO7v58GKiAskLQscDYwG/gLsFBG1p2KjgbGV044mV+t6iRwDv29E9BmTbmZmZtaBE4EPASdIeh+54tUDwMrA9mRvmvtoM8HjWMfMzMy6Yb6YZDkizgDOaHJsfOXnHwI/7EKzzMzM7D9QRDwuaUvgPGBjsrdOkBMjA1wP7BURjw/gmo51zMzMbFgNOMEjaRvyqdbREfFgg+MrAV8Fzo2IawfdQjMzM7Mui4jpwKaS1gM2Al4HPAHcGBG39rJtZmZmZo100oPnUGCtRskdgIh4UNLGwGuBawfRNjMzM7OeKskcJ3TMzMxsvrdAB+esR3ZN7s91wLgOrm1mZmY235C0pKR1JW3W67aYmZmZ9aeTBM8ooGHvnTqPlHJmZmZmI46kVSRdBDwOTAWuqTv2Tkl/LfP0mJmZmc0XOknwPAms2qLMqsDsDq5tZmZm1lOSRgM3AbsD/wfcwNwJlinHRgF7dr91ZmZmZo11kuC5GdhD0oqNDpZJlvco5czMzMxGmi+RCZx3RcS7gd/WH4yIF4DfA5v2oG1mZmZmDXWS4Pk28Brg95J2k7QogKRFJe0O/A5YCjh16JppZmZm1jU7Ab9qsRrofcBK3WmOmZmZWWsDXkUrIq6Q9BXgi8AvgJD0OPB6svuygC9HxG+GtKVmZmZm3bEC8PcWZV4AluxCW8zMzMza0kkPHiLiS8AOwK+Bx8gl0R8DLgW2j4gJQ9VAMzMzsy57jNbzDa4JPNyFtpiZmZm1paMED2RPnojYNSJGRcQi5XW3iPht67PnJekgSfdKmiNpWrtLkUpaQ9JTkp4e+DswMzMza2gKsFs/8w2uQT7ouqbR8SbnONYxMzOzYdVxgmeoSNoTmAgcD6wLXA9cJmm1FuctAvyUnPPHzMzMbKh8E1gMmCxpR2AJAElLlp8vAV4GTmrnYo51zMzMrBs6TvBI+m9JX5N0saQr6/aPkfQ+Sa9v81JHAJMi4uyIuCMiDgUeAg5scd7XgT8BP+voDZiZmZk1EBE3AfsDY8hl0j9dDv27/PwG4KMRcXubl3SsY2ZmZsOuowSPpC8DtwKfAXYFtqpc8yfAXm1cZxFgfeCKyqErgE36OW9nYBfgsAE13MzMzKwNEXEOsA65KujNwD/I2OcM4L8j4rx2ruNYx8zMzLplwAkeSe8HjgZ+C7wdOKH+eETcA0wFdmvjcssBCwKPVPY/AjQb9z4aOBvYOyKeaqO9+0uaKmnqzJkz22iSmZmZGUTE3yPi8IjYOCLWjIgNIuLQiLgLQNLybVzGsY6ZmZl1RSc9eA4D7gZ2j4g/Ac83KHMHsMYArhmVn9VgX82Pge9ExI1tXTjirIgYFxHjll++nTjMzMzMrDlJr5V0PNmrp12OdczMzGxYdZLgeStweUQ0SuzUPAis0Ma1ZgEv0fcJ1ij6Pumq2Rr4kqQXJb0IfB9Ysvy8fxt1mpmZmTUkaXVJ75a0q6QVKscWk/Q54B7gKNqLoxzrmJmZWVd0kuARuXJEf1YA5rS6UEkSTQO2rRzallxhopG3kkPDatsxwLPl356E0MzMzDry/9u7+2C56vqO4+8PAcSHShWChFrMCD4BopZAKxILxSBldMooStVaUnUoRKGCDxXLYHAUaqsoWCIGbaP4hH0Cp5QhxYIPBMVQHY0ChUoiCKZJS1GBEIjf/nE2erPZm9y79+7u3ez7NXPmZM/5nd9+l+T8+M73/M45SS6imZXz98AVwOoki1r7jgRuA94HPJbmrVhP316f5jqSJKlfdu7imNvZ9kMBZwFHABN9s8QFwGVJbgJuAE4B9gEuafV3PnBYVR0NUFWr2r5vHvCL9u2SJEkTleQk4C00F7Fuobmg9SzgoiQPAB+neZbOx4H3VdU9k+jeXEeSJPVcNwWeLwLvS/K2qvpQh/1nAfvTXNnarqq6PMkeNA9ungOsAo6rqjWtJnOA/bqIU5IkaaIW0jxX8KiquhEgyYtpXirxSeBu4OVV9b3JdmyuDZVHUwAAESNJREFUI0mS+qGbAs9HgFcBf5Xk1bQeEJjkg8B8YB7wDWDpRDusqiU0rx3ttG/hdo5dBiyb6HdJkiR1cDDwz5uLOwBV9dUkVwAnAG/oprgzpi9zHUmS1FOTLvBU1UNJjqKZofM6munKAGfSTGv+DPCWqnp02qKUJEnqrd1p3hLa7vbW+sYO+yRJkmaMbmbwUFX3AwuTnAkcCuwB3A/cVFXrpjE+SZKkftgJeKTD9kegucDV33AkSZImZ9IFniQ/BK6uqjdX1f8C10x/WJIkSX1Xgw5AkiSpW93M4JlNM1tHkiRpR7I4yeJOO5Js6rC5qqqr2dCSJEnTrZuk5Pv4pgdJkrTjSY/bS5Ik9cxOXRxzEfDyJAdPdzCSJEmDUFU7dbMMOm5JkqTNupnBczdwLXBDko8D3wJ+Qof71qvqq1MLT5IkSZIkSdvTTYHneppiTmhejb6tBxLO2sY+SZIkSZIkTYNuCjzvxbdMSJIkSZIkzRiTLvBU1eLpDiLJIuAdwByahzi/taq+Nk7bA4CLgQOA3YF7gC8Ai6tq43THJkmSNFXmOpIkqdcmVeBJsi9wKM0Mnm9V1V1TDSDJicCFwCLg66311UkOqKofdThkI/Ap4NvA/wHPAy6l+S3vnGo8kiRJ08lcR5Ik9cOECzxJPgi8lV+9ErSSfLiq3jHFGM4EllXVpa3PpyU5FjgVOKu9cVXdAdwxZtOaJEcC86cYhyRJUi+Y60iSpJ6b0Os9k7yWJjkJcCtwW+vPZyZ5TbdfnmRX4BBgeduu5cDhE+xjf+BY4CvdxiFJktQL5jqSJKlfJlTgAd4IPAq8pKoOrKoDgJcCv2jt69aeNG/aWtu2fS2w97YOTLIiyQbgdprpzu8ep93JSVYmWblu3bophCpJkjRp5jqSJKkvJlrgORi4oqqu27yhqq4FrgSePw1xtL+VKx22tTsR+C3gtcBxwJ937LhqaVXNq6p5s2fPnnKgkiRJXTDXkSRJPTXRZ/A8iea2rHa3AsdP4fvXA5vY+grWXmx9pWsLYx7w/IMks4BPJPnrqnp0CvFIkiRNJ3MdSZLUFxOdwbMT8EiH7Y/wq4cuT1rrVZ83Awvadi0AVkyiq51oilWzuo1FkiRpupnrSJKkfpnMa9K3N424WxcAlyW5CbgBOAXYB7gEIMn5wGFVdXTr8+uBDcD3aF4jOg84H/iHqnq4RzFKkiR1y1xHkiT13GQKPIuTLO60I8mmDpurqrbbf1VdnmQP4GxgDrAKOK6q1rSazAH2G3PIozSvFH0GzeyhNcDFwIcn+DskSZL6xlxHkiT1w2QKPJO9FWvC7atqCbBknH0L2z5/Hvj8JGORJEkaGHMdSZLUaxMq8FTVRJ/VI0mSJEmSpD6zcCNJkiRJkjTkLPBIkiRJkiQNOQs8kiRJkiRJQ84CjyRJkiRJ0pCzwCNJkiRJkjTkLPBIkiRJkiQNOQs8kiRJkiRJQ84CjyRJkiRJ0pCbEQWeJIuS3JlkQ5Kbk8zfRtsjk1yZ5N4kDyb5bpI39DNeSZKkyTDXkSRJvTbwAk+SE4ELgfOAFwArgKuT7DvOIYcD3wNOAA4CPgYsTfLaPoQrSZI0KeY6kiSpH3YedADAmcCyqrq09fm0JMcCpwJntTeuqvPaNn0syVHAK4HP9TRSSZKkyTPXkSRJPTfQGTxJdgUOAZa37VpOc/Vqop4I3DfOd5ycZGWSlevWresuUEmSpC6Y60iSpH4Z9C1aewKzgLVt29cCe0+kgyQvA44GlnbaX1VLq2peVc2bPXv2VGKVJEmaLHMdSZLUF4Mu8GxWbZ/TYdtWkryIZqry6VV1Uy8CkyRJmgbmOpIkqacGXeBZD2xi6ytYe7H1la4tJDkCuBo4p6o+1pvwJEmSpsRcR5Ik9cVACzxVtRG4GVjQtmsBzRsmOkryYpqE59yq+kjvIpQkSeqeuY4kSeqXmfAWrQuAy5LcBNwAnALsA1wCkOR84LCqOrr1+UjgKmAJ8Nkkm6+IbaoqnywoSZJmGnMdSZLUcwMv8FTV5Un2AM4G5gCrgOOqak2ryRxgvzGHLAQeB7y9tWy2Bpjb63glSZImw1xHkiT1w8ALPABVtYTmKlWnfQs7fF7Yqa0kSdJMZK4jSZJ6bdAPWZYkSZIkSdIUWeCRJEmSJEkachZ4JEmSJEmShpwFHkmSJEmSpCFngUeSJEmSJGnIWeCRJEmSJEkachZ4JEmSJEmShpwFHkmSJEmSpCE3Iwo8SRYluTPJhiQ3J5m/jba7JVmW5LtJHklyfR9DlSRJmjRzHUmS1GsDL/AkORG4EDgPeAGwArg6yb7jHDIL2AD8DXBVX4KUJEnqkrmOJEnqh4EXeIAzgWVVdWlV3VJVpwH3Aqd2alxVD1TVKVW1FLi7n4FKkiR1wVxHkiT13EALPEl2BQ4BlrftWg4c3v+IJEmSpo+5jiRJ6pdBz+DZk2Ya8tq27WuBvafjC5KcnGRlkpXr1q2bji4lSZImylxHkiT1xaALPJtV2+d02NZdx1VLq2peVc2bPXv2dHQpSZI0WeY6kiSppwZd4FkPbGLrK1h7sfWVLkmSpGFjriNJkvpioAWeqtoI3AwsaNu1gOYNE5IkSUPLXEeSJPXLzoMOALgAuCzJTcANwCnAPsAlAEnOBw6rqqM3H5DkAGBXmvvan5Dk+QBV9Z0+xy5JkrQ95jqSJKnnBl7gqarLk+wBnA3MAVYBx1XVmlaTOcB+bYf9K/C0MZ+/3Vqnl7FKkiRNlrmOJEnqh4EXeACqagmwZJx9Cztsm9vjkCRJkqaNuY4kSeq1QT9kWZIkSZIkSVNkgUeSJEmSJGnIWeCRJEmSJEkachZ4JEmSJEmShpwFHkmSJEmSpCFngUeSJEmSJGnIWeCRJEmSJEkachZ4JEmSJEmShpwFHkmSJEmSpCE3Iwo8SRYluTPJhiQ3J5m/nfbPTfKVJA8l+XGSc5KkX/FKkiRNhrmOJEnqtYEXeJKcCFwInAe8AFgBXJ1k33HaPxH4N2AtcChwOvAO4My+BCxJkjQJ5jqSJKkfBl7goUlWllXVpVV1S1WdBtwLnDpO+9cBjwNOqqpVVfWPwAeAM72yJUmSZiBzHUmS1HMDLfAk2RU4BFjetms5cPg4h70Q+FpVPTRm2zXAPsDc6Y5RkiSpW+Y6kiSpX3Ye8PfvCcyimYI81lrgJeMcszdwd4f2m/fdOXZHkpOBk1sff57ktq6j1bDaE1g/6CC0tXxg0BFoBHj+z1A9Pv+f1tPeJ8dcR/3gWDcDmeeoTzz/Z6hB5DqDLvBsVm2f02Hb9tp32k5VLQWWdh+ahl2SlVU1b9BxSOo/z3/NIOY66hnHOml0ef5rrEE/g2c9sInmatRYe7H1la7NfjJOe7ZxjCRJ0iCY60iSpL4YaIGnqjYCNwML2nYtoHnDRCc3AvOT7NbW/h5g9XTHKEmS1C1zHUmS1C+DnsEDcAGwMMmbkjwnyYU0DxG8BCDJ+Um+PKb954AHgWVJDkryCuBdwAVVta2pzhpdTluXRpfnv2YCcx31mmOdNLo8//VLmQl5QpJFwDuBOcAq4Iyq+mpr3zLgyKqaO6b9c4GLgcOA+2gSpPea9EiSpJnIXEeSJPXajCjwSJIkSZIkqXsz4RYtSZIkSZIkTYEFHs0oSeYmqdZ0dUkjxjFA0o7OcU4aXZ7/6jULPOq5JM9O8tEkq5Lcn2RjknuSXJXkjW1vCRlaSZ6a5G9bv+3hJKuTfCTJkwYdmzRIozAGJDmh9Ru/luSnreTtM4OOS1J/jMI4B+Y6UiejcP6b5wyPnQcdgHZsSc4B3kNTTPwG8Cng58BTgCOBTwCnAvMGFOK0SLIfzetu9wKuBG6leTDmnwHHJnlRVf3PAEOUBmJUxgDgbOB5NL/tbuDZgw1HUr+MyjhnriNtbVTOf8xzhoYFHvVMkncD5wJ3Aa+qqm92aPMy4G39jq0HltAkPKdX1Uc3b0xyAXAG8H7glAHFJg3EiI0BZ9AkPHcAvwtcN9hwJPXDiI1z5jrSGCN2/pvnDAlv0VJPJJkLLAYeAY7rNOABVNW/AMdOoL9nJvnLJCuTrGtNC16TZGmSp3ZonyQnJVnRar8hyV1JrklyYlvbg5N8vjXN+OFW+/9oTTneZQKxPR04BlhN80rbsd4DPAC8Psnjt9eXtKMYpTGg9Tuuq6rbfYW1NDpGaZwz15G2NErnf+t3mOcMCWfwqFf+BNgF+EJVrdpWw6p6eAL9vYLmqtB1NNODNwIHAm8CXp5kXlX9eEz79wNnAXcCXwTuB+YAhwKvAi6HZsADvgkU8KVW+ycC+wOLaKYjPrKd2H6vtV5eVb9o+20/S3IDTVL0O8CXJ/BbpR3BKI0BkkbTKI1z5jrSlkbp/NcQscCjXjmitZ6u/8lfBny4fYBMcgxwNc3gdOqYXX8K/Bg4qKoebDtmzzEfTwJ2A46vqivb2j0J2OLYcTyrtf7PcfbfTpP0PBOTHo2OURoDJI2mURrnzHWkLY3S+a8hYoFHvTKntb57Ojprq1iP3b48yfeBl3bY/QiwqcMx6zu0fahDu/smGN7urfX94+zfvP3XJ9iftCMYpTFA0mgapXHOXEfa0iid/xoiPoNHvZLWelru02zdZ/pHSa5t3Tf6aJrX8xXwXOA32g75LDAX+H6S85Mcm2T39n5ppi9uAq5I8ukkf5zmLRHTaVr/W0hDwjFA0o7OcW5M+K21uY5Ghee/ZiQLPOqVe1rrrR4K1qULaKYuHgBcA3yI5qn15wJrgF3b2p8BvJXmoX/vopnauD7JlUn239yoqm4C5gP/DpxA82rDO5LcmuQ1E4xt81WrToMqNPe5jm0njYJRGgMkjaZRGufMdaQtjdL5ryHiLVrqla/TPJDvaOCTU+koyV7A6cAq4PCq+lnb/q0Gp6raBFwIXNg6/gjgD2keOnZgkgM33+NaVTcCL0vyGOAQmifdnwZ8Lsm6qrp2OyHe1lo/c5z9z2itx7tvXdoRjdIYIGk0jdI4Z64jbWmUzn8NEWfwqFf+jua+0FcmOWBbDVuDzbY8nebf6vIOA95TW/vHVVX/XVX/VFWvpqle7wcc1KHdw1W1oqrOoRlkAf5gO7FB87R7gGOSbHFOJfk14EU0971+YwJ9STuKURoDJI2mURrnzHWkLY3S+a8hYoFHPVFVq4HFNNMJr0oyr1O7JMfSTCncltWt9RFJZo059gnApbTNREvymCRHJ0nb9l2AJ7c+PtjaNn+c+1WfMrbdtlTVfwHLae6DfXPb7nOBxwOfrqoHtteXtKMYpTFA0mgapXHOXEfa0iid/xou3qKlnqmq85LsDLwH+FaSFcBK4Oc0g8qLaab0rtxOPz9J8gWaaYffSbKc5h7wBcAG4DvA88cc8ljgWmB1km/S3Le6W6v9c4AvVdUtrbZvo7kadT3ww1ZsBwK/D9wHLJ3gz10ErAAuSnI0cAvw28BRNNOV/2KC/Ug7jFEaA5IcDxzf+rh3a/3CJMtaf15fVW+fSF+ShscojXOY60hbGKXz3zxniFSVi0tPF5qB5qM095X+FNgI3EtTzX4j8JgxbefSPI1+WVsfjwPeD9xBM9DdBVwM7AFc3/xT/mXbXYB3tvr/Uav9Opppw6cAu45pewzNFMsf0DwY8AGa+8wvAp42yd/5m62+7m39xjU098Y+edB/By4ug1xGYQyguYpX21hWD/rvwcXFpXfLKIxzrb7MdVxc2pZROP/Nc4ZnSesvTJIkSZIkSUPKZ/BIkiRJkiQNOQs8kiRJkiRJQ84CjyRJkiRJ0pCzwCNJkiRJkjTkLPBIkiRJkiQNOQs8kiRJkiRJQ84CjyRJkiRJ0pCzwCNJkiRJkjTkLPBIkiRJkiQNuf8HAsH1OSJ8H3UAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 1152x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "get_precisions_recalls(labels_test, preds)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Solution 2: Oversample Minority Class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "#get indices of each class\n",
    "indices_0 = np.where(labels_train == 0)[0]\n",
    "indices_1 = np.where(labels_train == 1)[0]\n",
    "indices = np.concatenate([indices_0, indices_1])\n",
    "\n",
    "#get weights for each class\n",
    "weights = np.empty(indices_0.shape[0] + indices_1.shape[0])\n",
    "weights[:indices_0.shape[0]] = 1\n",
    "weights[indices_0.shape[0]:] = weight_minority_class\n",
    "weights = weights/np.sum(weights)\n",
    "\n",
    "#sample new indices\n",
    "sampled_indices = np.random.choice(indices, indices.shape[0], p=weights)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "#consruct oversampled datasets\n",
    "X_train_oversampled = X_train[sampled_indices]\n",
    "labels_train_oversampled = labels_train[sampled_indices]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fraction of positive labels in oversampled data: 50.26%\n"
     ]
    }
   ],
   "source": [
    "print('Fraction of positive labels in oversampled data:', str(round(100*np.mean(labels_train_oversampled),3)) + '%')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf = DecisionTreeClassifier()\n",
    "clf.fit(X_train_oversampled, labels_train_oversampled)\n",
    "preds = clf.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "get_precisions_recalls(labels_test, preds)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Solution 2a: SMOTE (Synthetic Minority Oversampling TEchnique)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "#populate entire dataset\n",
    "X_train_oversampled_smote = []\n",
    "labels_train_oversampled_smote = []\n",
    "\n",
    "for _ in range(X_train.shape[0]):\n",
    "    p = np.random.random()\n",
    "    #sample from majority class\n",
    "    if p < 0.5:\n",
    "        X_train_oversampled_smote.append(X_train[np.random.choice(indices_0)])\n",
    "        labels_train_oversampled_smote.append(0)\n",
    "    #sample from minority class\n",
    "    else:\n",
    "        #get two random samples from minority class\n",
    "        minority_samp_1 = X_train[np.random.choice(indices_1)]\n",
    "        minority_samp_2 = X_train[np.random.choice(indices_1)]\n",
    "        \n",
    "        #get random proportion with which to mix them\n",
    "        prop = np.random.random()\n",
    "        \n",
    "        #generate synthetic sample from minority class\n",
    "        synthetic_minority_samp = prop*minority_samp_1 + (1-prop)*minority_samp_2\n",
    "        X_train_oversampled_smote.append(synthetic_minority_samp)\n",
    "        labels_train_oversampled_smote.append(1)\n",
    "        \n",
    "X_train_oversampled_smote = np.array(X_train_oversampled_smote)\n",
    "labels_train_oversampled_smote = np.array(labels_train_oversampled_smote)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fraction of positive labels in SMOTE oversampled data: 51.823%\n"
     ]
    }
   ],
   "source": [
    "print('Fraction of positive labels in SMOTE oversampled data:', str(round(100*np.mean(labels_train_oversampled_smote),3)) + '%')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Num Unique Class=1 Feature Vectors with Naive Oversample: 104\n",
      "Num Unique Class=1 Feature Vectors with SMOTE Oversample: 199\n"
     ]
    }
   ],
   "source": [
    "unique_naive_oversample = np.unique(X_train_oversampled[labels_train_oversampled == 1], axis=0).shape[0]\n",
    "unique_smote_oversample = np.unique(X_train_oversampled_smote[labels_train_oversampled_smote == 1], axis=0).shape[0]\n",
    "\n",
    "print('Num Unique Class=1 Feature Vectors with Naive Oversample:', unique_naive_oversample)\n",
    "print('Num Unique Class=1 Feature Vectors with SMOTE Oversample:', unique_smote_oversample)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf = DecisionTreeClassifier()\n",
    "clf.fit(X_train_oversampled_smote, labels_train_oversampled_smote)\n",
    "preds = clf.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "get_precisions_recalls(labels_test, preds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}