Allstate_Kaggle_Challenge.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Allstate Claims Severity Kaggle Challenge (spring 2018)\n",
    "\n",
    "### The Problem\n",
    "\n",
    "Allstate, an US-based insurance company, has for the past couple of years been developing automated\n",
    "methods of predicting the cost, and hence severity, of claims. In December 2016 it put forward\n",
    "a recruitment challenge on Kaggle to allow people to show off their creativity and present their\n",
    "technical skills by creating an algorithm which accurately predicts claims severity. With this the\n",
    "company hoped to get insight into better ways to predict claims severity and thereby ensure a\n",
    "more worry-free customer experience. This project takes on the 2016 Allstate challenge and\n",
    "use the provided datasets on Kaggle. The objective was to compare performance of different machine learning algorithms and methods on the insurance claim dataset from Allstate and to get the best model to predict the accident loss.\n",
    "\n",
    "### Outline\n",
    "\n",
    "First the data is observed, transformed and encoded.\n",
    "Then different prediction methods are used and compared. The methods that are used are \n",
    "- Linear Regression\n",
    "    - Vanilla LR\n",
    "    - Regularization\n",
    "        - Ridge\n",
    "        - Lasso\n",
    "- Support Vector Machines\n",
    "- Random Forest\n",
    "    - Little digression into feature exploration\n",
    "- Neural Network \n",
    "\n",
    "\n",
    "Finally the model that seems to give the best results is trained on the full training set and predictions are made to the submission test provided by Kaggle. \n",
    "\n",
    "\n",
    "### Results\n",
    "\n",
    "The model that had best results was (with no surprise) the neural network. The NN-model really stood out on the leaderboard with around 10% better performance the the second model (RF).\n",
    "\n",
    "| Model             | training MAE | test MAE |\n",
    "|-------------------|--------------|----------|\n",
    "| Linear Regression | 1299.18      | 1312.04  |\n",
    "| Ridge Regression  | 1286.57      | 1302.32  |\n",
    "| Lasso Regression  | 1286.35      | 1302.35  |\n",
    "| SVM               | 1296.36      | 1308.77  |\n",
    "| Random Forest     | 466.58       | 1251.67  |\n",
    "| Neural Network    | 1074.78      | 1179.82  |\n",
    "\n",
    "The submissions were divided into public (multi-submission) and private submission with the prediction MAE being the benchmark. The NN-model ended up scoring a private MAE of 1147.20 and a public MAE of 1138.11. Compared to the leading submission on Kaggle (private MAE of 1109.71), this is an acceptable result.\n",
    "\n",
    "### Conclusion\n",
    "\n",
    "What became extremely clear very soon in this project was how much impact the preprocessing of\n",
    "the data had on the final results for most of the models. It came clear how critical data processing\n",
    "is in all real life machine learning problems. Future steps would involve transformations and rescaling of the parameters. They worked well as they were but they could probably be massaged to enhance model performance. For the prediction models, the future steps would probably include a bigger process for hyperparameter optimization and a better way for validating (e.g. CV). Regarding the NN model, some methods for reducing the overfitting should definitely be applied (e.g. dropout or some regularization)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preprocess"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Libraries ---------------------------------\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn.utils import shuffle\n",
    "import warnings; \n",
    "warnings.filterwarnings(\"ignore\") # ignore warnings for the purpose of readability\n",
    "\n",
    "# Loading and splitting data --------------------------------------\n",
    "trainData = pd.read_csv(\"train.csv\")\n",
    "submissionData = pd.read_csv(\"test.csv\")\n",
    "trainData.drop('cont12',axis=1,inplace=True)\n",
    "submissionData.drop('cont12',axis=1,inplace=True)\n",
    "\n",
    "# copy of the data for one hot encoding later on ..\n",
    "OHE_TrainDataCopy = trainData\n",
    "OHE_submissionDataCopy = submissionData\n",
    "\n",
    "# Not interested in the id's (copy kept for submission)\n",
    "trainData.drop('id',axis=1,inplace=True)\n",
    "submissionDataIDs = submissionData['id']\n",
    "submissionData.drop('id',axis=1,inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Categorial encode data both by\n",
    "#   - Label encoding ([A,B,B,C] -> [1,2,2,3])\n",
    "#   - One-hot-encoding ([A,B,B,C] -> [A=[1,0,0,0], B=[0,1,1,0], C=[0,0,0,1]])\n",
    "class MultiColumnLabelEncoder:\n",
    "    def __init__(self,columns = None):\n",
    "        self.columns = columns # array of column names to encode\n",
    "\n",
    "    def fit(self,X,y=None):\n",
    "        return self # not relevant here\n",
    "\n",
    "    def transform(self,X):\n",
    "        '''\n",
    "        Transforms columns of X specified in self.columns using\n",
    "        LabelEncoder(). If no columns specified, transforms all\n",
    "        columns in X.\n",
    "        '''\n",
    "        output = X.copy()\n",
    "        if self.columns is not None:\n",
    "            for col in self.columns:\n",
    "                output[col] = LabelEncoder().fit_transform(output[col])\n",
    "        else:\n",
    "            for colname,col in output.iteritems():\n",
    "                output[colname] = LabelEncoder().fit_transform(col)\n",
    "        return output\n",
    "\n",
    "    def fit_transform(self,X,y=None):\n",
    "        return self.fit(X,y).transform(X)\n",
    "\n",
    "\n",
    "# If we're not careful, we'll have uneven amounts of catagories in our train and test datasets.\n",
    "# Handled in 3 steps:\n",
    "\n",
    "# *1. concat data\n",
    "Data = pd.concat([trainData, submissionData])\n",
    "\n",
    "# *2. Categorial encoding\n",
    "OneHotEncoded_Data = pd.get_dummies(Data)\n",
    "columns2encode = []\n",
    "for i in range(116):\n",
    "    columns2encode.append(list(Data)[i])\n",
    "Data = MultiColumnLabelEncoder(columns2encode).fit_transform(Data)\n",
    "\n",
    "# *3. splitting again\n",
    "#   *3.a one-hot-encoding\n",
    "OneHotEncoded_trainData = OneHotEncoded_Data[:trainData.shape[0]]\n",
    "OneHotEncoded_submissionData = OneHotEncoded_Data[trainData.shape[0]:]\n",
    "OneHotEncoded_submissionData.drop('loss',axis=1,inplace=True)\n",
    "#    *3.b label encoding ([A,B,B,C] -> [1,2,2,3])\n",
    "trainData = Data[:trainData.shape[0]]\n",
    "submissionData = Data[trainData.shape[0]:]\n",
    "submissionData.drop('loss',axis=1,inplace=True)\n",
    "\n",
    "# splitting into training, test and validation set 80/20 --------------------------------------\n",
    "#  *one-hot-encoding\n",
    "OneHotEncoded_y_trainData = OneHotEncoded_trainData['loss']\n",
    "OneHotEncoded_X_trainData = OneHotEncoded_trainData[OneHotEncoded_trainData.columns.difference(['loss'])]\n",
    "OneHotEncoded_X_train, OneHotEncoded_X_test, OneHotEncoded_y_train, OneHotEncoded_y_test = \\\n",
    "train_test_split(OneHotEncoded_X_trainData, OneHotEncoded_y_trainData, test_size=0.2, random_state=20)\n",
    "#  *label encoding\n",
    "y_trainData = trainData['loss']\n",
    "X_trainData = trainData[trainData.columns.difference(['loss'])]\n",
    "X_train, X_test, y_train, y_test = train_test_split(X_trainData, y_trainData, test_size=0.2, random_state=20)"
   ]
  },
  {
   "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"
    },
    {
     "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": [
    "# Log transformation for output variables --------------------------------------\n",
    "\n",
    "### A quick visualization shows how y values are lognormaly distributed\n",
    "plt.hist(y_train,20)\n",
    "plt.xlabel(\"Loss\")\n",
    "plt.show()\n",
    "\n",
    "### Log Transformation \n",
    "y_train_log = np.log(y_train)\n",
    "y_test_log = np.log(y_test)\n",
    "plt.hist(y_train_log,20)\n",
    "plt.xlabel(\"log(Loss)\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Reducing dimensions...\n",
      "Plotting PCA projection...\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": [
    "# Dimensionality reduction\n",
    "from sklearn.decomposition import PCA\n",
    "import matplotlib.cm as cm\n",
    "\n",
    "print(\"Reducing dimensions...\")\n",
    "pca = PCA(n_components=2).fit(OneHotEncoded_X_train)\n",
    "X2D = pca.transform(OneHotEncoded_X_train)\n",
    "\n",
    "print(\"Plotting PCA projection...\")\n",
    "plt.scatter(X2D[:,0],X2D[:,1],cmap=cm.Paired,c = y_train_log,s=10)\n",
    "plt.xlabel('PC 1')\n",
    "plt.ylabel('PC 2')\n",
    "plt.title('PCA projection')\n",
    "cbar = plt.colorbar()\n",
    "cbar.set_label(\"log-loss\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Modeling\n",
    "\n",
    "Few of the models that did best in predicting the output."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Split our training set further into train and validation\n",
    "X_train_, X_val_, y_train_log_, y_val_log_ = train_test_split(\n",
    "    X_train, y_train_log , test_size=0.2, random_state=20)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Regressions\n",
    "\n",
    "#### Vanila regression\n",
    "We'll start of by applying simple linear regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--------------------------------------------------------------\n",
      "Results from simple regression on the multi-labeled dataset:\n",
      "Training MAE using log transformation: 1299.1844015222082\n",
      "Test MAE using log transformation: 1312.04026607101\n",
      "--------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "### Simple Regression ###\n",
    "from sklearn import linear_model\n",
    "\n",
    "# Multi-labeled dataset with log(Loss) as an output\n",
    "reg = linear_model.LinearRegression(fit_intercept=False, normalize=True)\n",
    "reg.fit(X_train, y_train_log)\n",
    "y_pred_train_log = reg.predict(X_train)\n",
    "y_pred_test_log = reg.predict(X_test)\n",
    "print('--------------------------------------------------------------')\n",
    "print(\"Results from simple regression on the multi-labeled dataset:\")\n",
    "print('Training MAE using log transformation:', \n",
    "      np.average(abs(np.exp(y_train_log)-np.exp(y_pred_train_log))))\n",
    "print('Test MAE using log transformation:', \n",
    "      np.average(abs(np.exp(y_test_log)-np.exp(y_pred_test_log))))\n",
    "print('--------------------------------------------------------------')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Shrinkage methods\n",
    "\n",
    "Regularizations that penalize for bigger number of parameters to avoid overfitting. \n",
    "Ridge Rigression and the Lasso."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "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"
    },
    {
     "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": [
      "--------------------------------------------------------------------------------------\n",
      "Results from simple regression on the multi-labeled dataset:\n",
      "Training MAE using log transformation: 1286.5653087229948\n",
      "test MAE using log transformation: 1302.3190531806285\n",
      "--------------------------------------------------------------------------------------\n",
      "--------------------------------------------------------------------------------------\n",
      "Results from simple regression on the multi-labeled dataset:\n",
      "Training MAE using log transformation: 1286.351826057454\n",
      "test MAE using log transformation: 1302.3492578758494\n",
      "--------------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "def plot_hyp_param_search(alphas, train_MAE, val_MAE, title):\n",
    "    plt.plot(np.log10(alphas),train_MAE,label='training MAE')\n",
    "    plt.plot(np.log10(alphas),val_MAE,label='validation MAE')\n",
    "    plt.xlabel('log($\\lambda$)')\n",
    "    plt.ylabel('MAE')\n",
    "    plt.title(title)\n",
    "    plt.legend()\n",
    "    plt.show()\n",
    "    \n",
    "def hyp_param_search(model, alphas, X_train_, y_train_log_, X_val_, y_val_log_, name):\n",
    "    train_MAE = np.zeros(alphas.shape[0])\n",
    "    val_MAE = np.zeros(alphas.shape[0])\n",
    "    for a in range(alphas.shape[0]):\n",
    "        reg = model(alpha = alphas[a])\n",
    "        reg.fit(X_train_, y_train_log_)\n",
    "        y_pred_train_log = reg.predict(X_train_)\n",
    "        y_pred_val_log = reg.predict(X_val_)\n",
    "        train_MAE[a] = np.average(abs(np.exp(y_train_log_)-np.exp(y_pred_train_log)))\n",
    "        val_MAE[a] = np.average(abs(np.exp(y_val_log_)-np.exp(y_pred_val_log)))\n",
    "    \n",
    "    plot_hyp_param_search(alphas, train_MAE, val_MAE, \n",
    "                      title = \"Hyperparameter search for \" + name + \" regression\")\n",
    "    \n",
    "    return alphas[np.argmin(val_MAE)]\n",
    "\n",
    "def eval_perf(reg, X_train, y_train_log, X_test, y_test_log):\n",
    "    print('--------------------------------------------------------------------------------------')\n",
    "    reg.fit(X_train, y_train_log)\n",
    "    y_pred_train_log = reg.predict(X_train)\n",
    "    y_pred_test_log = reg.predict(X_test)\n",
    "    print(\"Results from simple regression on the multi-labeled dataset:\")\n",
    "    print('Training MAE using log transformation:', \n",
    "          np.average(abs(np.exp(y_train_log)-np.exp(y_pred_train_log))))\n",
    "    print('test MAE using log transformation:', \n",
    "          np.average(abs(np.exp(y_test_log)-np.exp(y_pred_test_log))))\n",
    "    print('--------------------------------------------------------------------------------------')\n",
    "\n",
    "# Hyperparameter search\n",
    "alphas = 1/np.power(10,np.arange(5))\n",
    "alpha_Ridge = hyp_param_search(linear_model.Ridge, alphas, X_train_, y_train_log_,\n",
    "                               X_val_, y_val_log_, name=\"Ridge\")\n",
    "alpha_Lasso = hyp_param_search(linear_model.Lasso, alphas, X_train_, y_train_log_, \n",
    "                               X_val_, y_val_log_, name=\"Lasso\")\n",
    "\n",
    "# Evaluating the performance of our best hyperparameter on the test set\n",
    "ridgeReg = linear_model.Ridge(alpha = alpha_Ridge)\n",
    "eval_perf(ridgeReg, X_train, y_train_log, X_test, y_test_log)\n",
    "lassoReg = linear_model.Ridge(alpha = alpha_Lasso)\n",
    "eval_perf(lassoReg, X_train, y_train_log, X_test, y_test_log)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### SVM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEHCAYAAABfkmooAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAgAElEQVR4nO3dd3hVVdbA4d9KJw0CCaETwAASOqEoRUAErEgRgg1GHRRlHB3FOp+gjjOOYp0RRwVEBQKIoqgIyCiDICpBioCUCBFCCaEYEgIhZX9/nMvlhpSbhFtS1vs89zF3n31O1gXJyj5773XEGINSSilVGh9vB6CUUqry02ShlFLKKU0WSimlnNJkoZRSyilNFkoppZzy83YA7hAZGWliYmK8HYZSSlUpGzZsOGqMiSruWLVMFjExMSQlJXk7DKWUqlJE5LeSjultKKWUUk5pslBKKeWUJgullFJOVcs5C6WU5+Tm5pKamsqZM2e8HYoqo6CgIJo0aYK/v3+Zz9FkoZS6KKmpqYSFhRETE4OIeDsc5YQxhmPHjpGamkqLFi3KfJ7bbkOJyCwROSIiWx3aOovI9yKySUSSRKSHrV1E5HURSRaRLSLS1eGccSKy2/Ya5654lVIVc+bMGerVq6eJoooQEerVq1fukaA75yxmA0MvaHsBeNoY0xl4yvYe4Gog1vaaALwJICJ1gSlAT6AHMEVEItwYs1KqAjRRVC0V+ftyW7IwxqwGjl/YDITbvq4NHLR9PQx431i+B+qISENgCPCVMea4MeYE8BVFE5DLFBQYZny7h+OnzrrrWyilVJXk6dVQDwAvish+YBrwuK29MbDfoV+qra2k9iJEZILt1lZSenp6hYLbe+wULyzfyd0fJJGTl1+hayilPOv3339n+vTpFTr3mmuu4ffffy+1z1NPPcXKlSsrdP0LxcTE0Ldv30JtnTt3pn379oXa/vznP9O4cWMKCgrsbbNnzyYqKorOnTvbX9u3b3dJXGXh6WQxEXjQGNMUeBCYaWsvbkxkSmkv2mjM28aYeGNMfFRUsbvVnWoVFcpLN3VifcoJHvvoZ/TBUEpVfqUli/z80n/pW7p0KXXq1Cm1zzPPPMOgQYMqHN+FMjMz2b/f+h34l19+KXK8oKCAxYsX07RpU1avXl3o2JgxY9i0aZP91a5dO5fF5Yynk8U44GPb1x9izUOANWJo6tCvCdYtqpLa3eb6To146KrWLN54gH9/nezOb6WUcoHHHnuMX3/9lc6dOzN58mRWrVrFgAEDuPnmm+nQoQMAN954I926dSMuLo63337bfm5MTAxHjx4lJSWFSy+9lD/+8Y/ExcUxePBgTp8+DcD48eNZtGiRvf+UKVPo2rUrHTp0YMeOHQCkp6dz1VVX0bVrV+6++26aN2/O0aNHi4139OjRLFiwAIDExETGjh1b6Pg333xD+/btmThxIomJia79w7oInl46exC4AlgFDAR229qXAJNEZD7WZHaGMeaQiCwH/u4wqT2Y87eu3GbSwEvYe/QUL321ixZRIVzXsZG7v6VS1cLTn21j+8GTLr1mu0bhTLk+rsTjzz//PFu3bmXTpk0ArFq1ih9//JGtW7fal4bOmjWLunXrcvr0abp3787IkSOpV69eoevs3r2bxMRE3nnnHUaPHs1HH33ErbfeWuT7RUZG8tNPPzF9+nSmTZvGjBkzePrppxk4cCCPP/44y5YtK5SQLjRq1CjGjx/Pww8/zGeffcbcuXP54IMP7MfPJZBhw4bxxBNPkJuba98PsWDBAtasWWPvu27dOmrVqlWGP8WL586ls4nAOqCNiKSKyJ3AH4GXRGQz8HeslU8AS4E9QDLwDnAvgDHmOPAssN72esbW5lYiwj9GdqB7TAQPLdzMxn0n3P0tlVIu1KNHj0J7CF5//XU6depEr1692L9/P7t37y5yTosWLejcuTMA3bp1IyUlpdhrjxgxokifNWvWkJCQAMDQoUOJiCh50WbdunWJiIhg/vz5XHrppQQHB9uPnT17lqVLl3LjjTcSHh5Oz549WbFihf34hbehPJUowI0jC2PM2BIOdSumrwHuK+E6s4BZLgytTAL9fHnrtnhufGMtf3w/iU/u602TiGDnJypVg5U2AvCkkJAQ+9erVq1i5cqVrFu3juDgYPr371/sHoPAwED7176+vvbbUCX18/X1JS8vD6Dc85tjxozhvvvuY/bs2YXaly1bRkZGhv32WXZ2NsHBwVx77bXlur47aG2oUtQNCWDW+O7k5BVw5+wkMs/kejskpdQFwsLCyMzMLPF4RkYGERERBAcHs2PHDr7//nuXx9CnTx8WLlwIwIoVKzhxovS7EcOHD+eRRx5hyJAhhdoTExOZMWMGKSkppKSksHfvXlasWEF2drbLYy4vTRZOXFI/lDdv6UZyehaT5m0kL7/A+UlKKY+pV68evXv3pn379kyePLnI8aFDh5KXl0fHjh35v//7P3r16uXyGKZMmcKKFSvo2rUrX375JQ0bNiQsLKzE/mFhYTz66KMEBATY27Kzs1m+fHmhUURISAh9+vThs88+A6w5C8els999953LP0tJpDouD42PjzeufvjRvB/28cTinxl3WXOeHtbe+QlK1RC//PILl156qbfD8KqcnBx8fX3x8/Nj3bp1TJw40T7hXlkV9/cmIhuMMfHF9ddCgmV0c89m7EnPYsaavbSMCmXc5THeDkkpVUns27eP0aNHU1BQQEBAAO+88463Q3I5TRbl8Pg1l5JyLJunP9tGs3rBDGhT39shKaUqgdjYWDZu3OjtMNxK5yzKwddHeC2hM20bhPOneRvZebjkSTWllKpONFmUU0igHzPHxxMS6Msds9eTnpnj7ZCUUsrtNFlUQMPatZhxe3eOnzrLH99P4kyuFh1USlVvmiwqqEOT2rwypjObU3/noQ83U1BQ/VaVKaXUOZosLpSaBPl5Zeo6tH0DHh3ali+2HOKVlbvcHJhSylVCQ0MBOHjwIKNGjSq2T//+/XG2BP/VV18ttGGuLCXPy2Lq1KmICMnJ54uZvvLKK4hIoZg2btyIiLB8+fJC5/v6+hbaj/H8889fdEyaLByl74KZg+HLR6CM+0/u7teS0fFN+NfXyXz8U6qbA1RKuVKjRo3sFWUr4sJkUZaS52XVoUMH5s+fb3+/aNGiIiXJExMT6dOnT5HqtLVq1SpUQ+qxxx676Hg0WTiKag2974ekmbD2tTKdIiL87cYOXNayHo999DPrU9xe51Ap5eDRRx8t9DyLqVOn8tJLL5GVlcWVV15pLyf+6aefFjk3JSXF/uCh06dPk5CQQMeOHRkzZkyh2lATJ04kPj6euLg4pkyZAljFCQ8ePMiAAQMYMGAAcL7kOcDLL79M+/btad++Pa+++qr9+5VUCv1CN954oz3mPXv2ULt2bRyf1WOMYdGiRcyePZsVK1aU+5na5aX7LC408Cn4fT+snAK1m0CH4oeojgL8fHjz1q6MmP4dE2xFB5vXC3F6nlLVzpePweGfXXvNBh3g6pJvoyQkJPDAAw9w7733ArBw4UKWLVtGUFAQixcvJjw8nKNHj9KrVy9uuOGGEp8//eabbxIcHMyWLVvYsmULXbt2tR977rnnqFu3Lvn5+Vx55ZVs2bKF+++/n5dffplvvvmGyMjIQtfasGED7777Lj/88APGGHr27MkVV1xBREREmUuhh4eH07RpU7Zu3cqnn37KmDFjePfdd+3H165dS4sWLWjVqhX9+/dn6dKl9oq4p0+ftlfQBXj88ccZM2ZMGf6wS6Yjiwv5+MCN06F5H/hkIqSsLdNpdYIDmDm+Owa4Y/Z6Mk5r0UGlPKFLly4cOXKEgwcPsnnzZiIiImjWrBnGGJ544gk6duzIoEGDOHDgAGlpaSVeZ/Xq1fYf2h07dqRjx472YwsXLqRr16506dKFbdu2OX2c6Zo1axg+fDghISGEhoYyYsQIvv32W6DspdDBSoTz58/nk08+Yfjw4YWOJSYm2suiJyQkFLoVdeFtqItNFKAji+L5BULCHJg5BOaPhTu/gqg2Tk9rERnCW7d249aZP3Dv3A3M/kMP/H01H6sapJQRgDuNGjWKRYsWcfjwYfsP0Llz55Kens6GDRvw9/cnJibG6a2a4kYde/fuZdq0aaxfv56IiAjGjx/v9Dql1dwrayl0gOuvv57JkycTHx9PeHi4vT0/P5+PPvqIJUuW8Nxzz2GM4dixY2RmZpZawPBi6E+yktSKgFs+BN9AmDsKMkv+jcRRz5b1+MeIjqxNPsZTn27V53gr5QHnfgNftGiRfXVTRkYG9evXx9/fn2+++Ybffvut1Gv069ePuXPnArB161a2bNkCwMmTJwkJCaF27dqkpaXx5Zdf2s8pqTx6v379+OSTT8jOzubUqVMsXryYvn37lvtz1apVi3/+8588+eSThdpXrlxJp06d2L9/PykpKfz222+MHDmSTz75pNzfo6w0WZQmojncshBOHYV5o+HsqTKdNqpbE+7t34rEH/cz49u9bg5SKRUXF0dmZiaNGzemYcOGANxyyy0kJSURHx/P3Llzadu2banXmDhxIllZWXTs2JEXXniBHj16ANCpUye6dOlCXFwcd9xxB71797afM2HCBK6++mr7BPc5Xbt2Zfz48fTo0YOePXty11130aVLlwp9toSEhELzJ2DdgrrwttTIkSOZN28ecH7O4tzLFauhtER5WexaDokJEDsYxswFX+d37woKDJMSf+LLrYd569ZuDI5r4Lp4lKpEtER51VTeEuU6siiL1kPg2pdg17Iy78Hw8RFeuqkzHRvX5s/zN7H1QIYHAlVKKffQZFFW8XdAnwfLtQejVoAv79weT0SwP3e9l8ThDPeug1ZKKXfRZFEeA5+C9qOsPRg/l23XZ/3wIGaM607mmVzuen892WfLVkpEqaqkOt7Ors4q8velyaI8KrgHo12jcP51cxe2HzzJA/M3adFBVa0EBQVx7NgxTRhVxLlltkFBQeU6Tye4K+L0CWsPRtbhMu/BAJi1Zi/PfL6du/u15PFrdEJQVQ+5ubmkpqa6vdyEcp2goCCaNGmCv79/oXZ9BrernduDMWMQzBkFd62EsGinp/2hdwx7jmbx1uo9tIgMIaFHMw8Eq5R7+fv706JFC2+HodxMb0NV1Lk9GNlHYd5NkJPl9BQRYer1cfSNjeSvn2zlu+SjHghUKaUuniaLi9GoC9w02yqctuiOMj0Hw8/Xhzdu6UqLyBDumbOBX9OdJxmllPI2TRYX69wejN3L4cvJZdqDER7kz6zx3fH39eGO2es5ceqsBwJVSqmK02ThCvY9GLNg7atlOqVp3WDevr0bhzLOcPecDeTk6XO8lVKVlyYLV7HvwZha5j0Y3ZrX5cVRHflx73Ge+FiLDiqlKi9dDeUq5/ZgZB629mCENYCYPk5PG9a5MXuPnuLVlbtpGRXCfQMu8UCwSilVPm4bWYjILBE5IiJbHdoWiMgm2ytFRDY5HHtcRJJFZKeIDHFoH2prSxaRiy+d6E7nnoMR0QLm3wzpO8t02p+vjGVY50a8uHwnX2w55OYglVKq/Nx5G2o2MNSxwRgzxhjT2RjTGfgI+BhARNoBCUCc7ZzpIuIrIr7AG8DVQDtgrK1v5eX4HIw5ZXsOhojwz5Ed6dY8gr8s3MSm/b97IFCllCo7tyULY8xq4Hhxx8R6HNVo4NxzAIcB840xOcaYvUAy0MP2SjbG7DHGnAXm2/pWbhXYgxHk78tbt3UjKiyQu95L4sDvJT89SymlPM1bE9x9gTRjzG7b+8bAfofjqba2ktqLEJEJIpIkIknp6eluCLmcKrAHIzI0kHfHdycnN587Z68nK0eLDiqlKgdvJYuxnB9VABR98C2YUtqLNhrztjEm3hgTHxUV5YIQXaACezBio8N445au7D6Sxf2JG8nXooNKqUrA48lCRPyAEcACh+ZUoKnD+ybAwVLaq44K7MHo1zqKqTfE8fWOI/zti+1uDlAppZzzxtLZQcAOY0yqQ9sSYJ6IvAw0AmKBH7FGFrEi0gI4gDUJfrOH4714A5+C3/dZezBqN4UOo5yecluv5uxJz+LdtSm0jAzhtsti3B6mUkqVxG3JQkQSgf5ApIikAlOMMTOxfuA73oLCGLNNRBYC24E84D5jTL7tOpOA5YAvMMsYs81dMbuNjw/c+Ga592D89dp2/HYsm6mfbadZvRCuaF1Jbq8ppWocfZ6FJ2Ufh1lDICutzM/ByMrJY9Sb33HgxGk+uvdyWkeHeSBQpVRNVNrzLLTchycF1y33HozQQD9mje9OUIAvd8xez9GsHA8EqpRShWmy8LSIGLh5Qbn2YDSqU4sZt8dzNCuHCe8ncSZXiw4qpTxLk4U3NO4Ko94t1x6MTk3r8PLozvy073ceWbRFiw4qpTxKk4W3tBkK10wr1x6Mazo0ZPKQNizZfJBXV+522l8ppVxFq856U/c7rSW1a1+FOs2s/RhO3Nu/FXvST/Haf3fTIjKEG7sUu6FdKaVcSpOFt105BTL2l3kPhojwjxEd2H8im0cWbaFJRC3iY+p6JlalVI2lt6G87dwejOa9rT0YKWucnhLg58Nbt3ajUZ0gJnywgX3Hsj0QqFKqJtNkURn4BcKYOdZKqTI+ByMiJICZ47uTX2C44731ZJzOdX+cSqkaS5NFZVGBPRitokJ589aupBw9xaR5P5GbX+CBQJVSNZEmi8qkAnswLm8Vyd+Hd+Db3UeZumSbLqlVSrmFJovKpgJ7MEZ3b8rdV7Rk7g/7mLU2xf0xKqVqHE0WlVEF9mA8OqQtQ+Ki+dsX21m53fktLKWUKg9NFpVV9zuh9wNlfg6Gj4/wypjOxDUK5/75G9l2MMMDQSqlagpNFpXZlVOg/UhrD8bPi5x2Dw7wY+a47oQH+XPXe0kcOXnG/TEqpWoETRaVWQX2YESHBzFjXDy/Z+dy1/tJnD6rRQeVUhdPk0VlV4E9GO0b1+b1sV34+UAGDy7YRIE+x1spdZE0WVQFFdiDcVW7aJ685lKWbTvMiyucJxillCqNJouqogJ7MO7s04KxPZrx5qpfWZi03/0xKqWqLU0WVUk592CICM8Mi6PPJZE8ufhn1v16zEOBKqWqG00WVU0592D4+/rwxi1daVY3mHvmbGBPuvMRiVJKXUiTRVVUzj0YtWv5M2t8d3wE7nwvid+zz3ogSKVUdaLJoqoq5x6M5vVCePv2eA6cOM3dH2zgbJ4WHVRKlZ0mi6qqAnswusfU5Z+jOvDD3uM8ufhnLTqolCozTRZVWQX2YAzv0oQ/DbyEDzek8p//7XF/jEqpakGTRVVXgT0YDw5qzXUdG/LPZTtYtvWQB4JUSlV1miyqg3LuwfDxEabd1IkuzerwwIJNbEn93TNxKqWqLE0W1UU592AE+fvy9m3x1AsJ5M73kjicoUUHlVIl02RRnZRzD0ZUWCCzxncn60weDyzYSL7WkFJKlUCTRXVTzj0YbRqE8fSwOL7fc5w3VyV7IEClVFWkyaI6KucejJu6NeH6To14ZeVuNvx23P3xKaWqHLclCxGZJSJHRGTrBe1/EpGdIrJNRF5waH9cRJJtx4Y4tA+1tSWLyGPuirdaKeceDBHhueHtaVg7iPsTN5FxOtdDgSqlqgp3jixmA0MdG0RkADAM6GiMiQOm2drbAQlAnO2c6SLiKyK+wBvA1UA7YKytr3KmnHswwoP8eX1sFw6fPMMTH+uGPaVUYW5LFsaY1cCF9zQmAs8bY3JsfY7Y2ocB840xOcaYvUAy0MP2SjbG7DHGnAXm2/qqsijnHoyuzSJ4aHBrvvj5EAvWa0lzpdR5np6zaA30FZEfROR/ItLd1t4YcPzplGprK6m9CBGZICJJIpKUnp7uhtCrqHLuwbinXyt6X1KPqZ9tI/lIpmdiVEpVep5OFn5ABNALmAwsFBEBpJi+ppT2oo3GvG2MiTfGxEdFRbkq3uqhHHswfHyEl0d3JjjAj0nzNnImV5/hrZTyfLJIBT42lh+BAiDS1t7UoV8T4GAp7aq8yrEHIzo8iGk3dWTH4Uye/3KHB4NUSlVWnk4WnwADAUSkNRAAHAWWAAkiEigiLYBY4EdgPRArIi1EJABrEnyJh2OuPsqxB2Ng22j+0DuG2d+lsHK783pTSqnqzZ1LZxOBdUAbEUkVkTuBWUBL23La+cA42yhjG7AQ2A4sA+4zxuQbY/KAScBy4Bdgoa2vqqhy7MF47Oq2tGsYzuRFm7UciFI1nFTHJZLx8fEmKSnJ22FUXnk58MFwSF0Pty2GmD4ldv01PYvrXl9Dp6a1mXtXL3x9iptGUkpVByKywRgTX9wx3cFdE5VjD0arqFAtB6KU0mRRY124ByPrSIldtRyIUkqTRU12bg/GqXRYeDvknS2227lyII3qaDkQpWoqTRY1XeOuMOzfsG8dLCu59FZ4kD+vJ3QhTcuBKFUjabJQ0GEUXH4/JM2EDbNL7NalWQR/0XIgStVImiyUZdBUaDUQvngY9v1QYjctB6JUzaTJQll8fGHkTKjdGBbeBicPFd9Ny4EoVSOVmixEJLyUY81cH47yquC6kJBoFRtccKu1H6MYWg5EqZrH2chi1bkvROS/Fxz7xOXRKO+LbgfD/wMHkuCLv5RYQ0rLgShVszhLFo7bdeuWckxVJ+1ugH6TYeMcWD+jxG5aDkSpmsNZsjAlfF3ce1Wd9H8CWg+1ltOmrC22S6CfL/+6uQtncgt4YMFG8gv0fwmlqitnyaK+iPxFRB5y+Prce31oRHXm4wMj3oaIFtaGvYzUYrtpORClagZnyeIdIAwIdfj63PuS70+o6iGoNiTMsya6598CuaeL7ablQJSq/ipcdVZEuhtj1rs4HpfQqrMutvNLSEyAjmNg+FsgRaerTp7J5drXv6WgAJb+uS+1a/l7IVCl1MVwWdVZEWknIs+IyG7gTZdEpyq/NlfDgCdhywL4fnqxXbQciFLVm9NkISLNReQxEdkMfADcC1xVUvZR1VTfh6HtdbDir/DrN8V20XIgSlVfzjblfQcsBfyBUcaYbkCmMSbFA7GpysTHx9p/EdkaFv0BTqQU203LgShVPTkbWaRjTWhHc371k95fqKkCw6wJb1NgTXifPVWki5YDUap6KjVZGGOGAR2An4CnRWQvECEiPTwRnKqE6rWCkbMgbRt8el+xO7y1HIhS1Y/TOQtjTIYxZpYx5iqgFzAFeFVE9KZ0TRU7CAZNgW2LYe2rxXYZ2DaaO3q30HIgSlUT5VoNZYxJM8a8boy5HOjjpphUVdD7AYgbDiufht0ri+3y6NVttByIUtVEqfssRGRJaScbY25weUQuoPssPOTsKZg5GDL2wx+/sW5RXeDX9Cyue30NnZrWZu5dvfD10ZJiSlVWF7PP4jKgCfAtMA146YKXqskCQiBhLogPzL8ZcoquftJyIEpVD86SRQPgCaA98BpwFXDUGPM/Y8z/3B2cqgIiYuCm2XB0Fyy+BwoKinTRciBKVX3OVkPlG2OWGWPGYU1uJwOrRORPHolOVQ0t+8Pgv8GOz+HbaUUOiwjPDW9PozpB3J+4iYzTuR4PUSl1ccqygztQREYAc4D7gNeBj90dmKpiet1r1Y765jnYsbTIYS0HolTV5mwH93vAd0BX4GljTHdjzLPGmAMeiU5VHSJw/WvQsBN8PAHSdxXpouVAlKq6nI0sbgNaA38GvhORk7ZXpoicdH94qkrxrwVj5oJfIMwfC2cyinTRciBKVU3O5ix8jDFhtle4wyvMGBPuqSBVFVKnKYx+36od9dEfi0x4azkQpaqmcm3KKw8RmSUiR0Rkq0PbVBE5ICKbbK9rHI49LiLJIrJTRIY4tA+1tSWLyGPuile5UExvGPo87F4Oq/5e5HB0eBAv3dRJy4EoVYW4LVkAs4GhxbS/YozpbHstBes5GUACEGc7Z7qI+IqIL/AGcDXQDhhr66squ+53QZdbYfWLsP3TIocHtK2v5UCUqkLcliyMMauBsi6qHwbMN8bkGGP2Yi3R7WF7JRtj9hhjzgLzbX1VZScC174MjeNh8URI216ki5YDUarqcOfIoiSTRGSL7TZVhK2tMeC4PCbV1lZSu6oK/AJhzBwIDLUmvLML/+4Q6OfLv27uwpncAh5YsJH8Al1Oq1Rl5elk8SbQCugMHOJ8yZDiCgaZUtqLEJEJIpIkIknp6emuiFW5QnhDK2FkHICP7oSCwhPaWg5EqarBo8nCVrU23xhTALyDdZsJrBFDU4euTYCDpbQXd+23jTHxxpj4qKio4roob2naA659CX79Gv77dJHDN3Vrwg1aDkSpSs2jyUJEGjq8HQ6cWym1BEiw7RZvAcQCPwLrgVgRaSEiAViT4KVWwlWVVLdxEH8nrH0Nfl5U6JCI8DctB6JUpebOpbOJwDqgjYikisidwAsi8rOIbAEGAA8CGGO2AQuB7cAy4D7bCCQPmAQsB34BFtr6qqpo6PPQ7DL4dBIc2lLokJYDUapyK/V5FlWVPs+iEss6Am9dAT5+MGEVhNQrdHj6qmReWLaTf47swJjuzbwSolI11cU8z0Ip1wqtDwlzICsNPhwH+XmFDp8rBzJliZYDUaoy0WShPK9xN6voYMq38NX/FTrk4yO8ouVAlKp0NFko7+g8FnpOhO+nw6bEQofqazkQpSodTRbKewY/CzF94bM/w4GfCh3SciBKVS6aLJT3+Ppbj2QNjYYFt1qT3w60HIhSlYcmC+VdIZHWhHf2cVh4O+SdtR/SciBKVR6aLJT3NewEw/4N+9bB8scLHdJyIEpVDposVOXQYRRcfj+snwEb3it0SMuBKOV9mixU5TFoKrQaCF88BPt+sDdrORClvE+Thao8fHxh5Eyo3RgW3gYnD9kPaTkQpbxLk4WqXILrQsI8yMmyVkjl5dgPdWkWwV8Gt+aLnw+xMGl/KRdRSrmaJgtV+UTHwfA34UASfPEXcBhFaDkQpbxDk4WqnNoNg74Pw8Y51qS3jZYDUco7NFmoymvAExA7BJY9Bilr7c1aDkQpz9NkoSovH18Y8TZExFgb9jJS7Ye0HIhSnqXJQlVutepYE955OTD/Fsg9bT+k5UCU8hxNFqryi4isJigAABOySURBVGpjjTAObbKKDtomvM+VA8nJK+DBBZu0HIhSbqTJQlUNba+B/k/AlgXw/Zv25lZRoTx9Qxzr9hzTciBKuZEmC1V19JsMba+DFX+FPavszaO0HIhSbqfJQlUdPj4w/D8QGQsf/gFOpABaDkQpT9BkoaqWwDBrwtvkw/xb4ewpQMuBKOVumixU1VOvFYycBWlb4dNJ9glvLQeilPtoslBVU+wgGDQFtn0Ma1+zN9/TrxV9LonUciBKuZgmC1V19X4A4obDyqmweyVglQN5eXQnLQeilItpslBVlwgMe8MqPPjRHXDsV0DLgSjlDposVNUWEAIJc0F8rB3eOdatJy0HopRrabJQVV9EDNw0G47uhMX3QEEBYJUDiWuk5UCUcgVNFqp6aNkfBv8NdnwO374E2MqBjNVyIEq5giYLVX30uhc6joFvnoOdXwLQUsuBKOUSmixU9SEC178GDTvCxxMgfReg5UCUcgW3JQsRmSUiR0RkazHHHhYRIyKRtvciIq+LSLKIbBGRrg59x4nIbttrnLviVdWEfy0YMxd8A2D+zXAmQ8uBKOUC7hxZzAaGXtgoIk2Bq4B9Ds1XA7G21wTgTVvfusAUoCfQA5giIhFujFlVB3Wawuj34cRea4RRUKDlQJS6SG5LFsaY1UBxY/5XgEcAx3+tw4D3jeV7oI6INASGAF8ZY44bY04AX1FMAlKqiJjeMPR52LUMVv0DsMqBPDS4jZYDUaoCPDpnISI3AAeMMZsvONQYcPzXm2prK6m9uGtPEJEkEUlKT093YdSqyup+F3S5FVa/ANuXAHB3v5b2ciDTVyWTfTbPy0EqVTV4LFmISDDwJPBUcYeLaTOltBdtNOZtY0y8MSY+Kiqq4oGq6kMErnkJGsdb+y/StuPjI7wypjO9W0XywrKd9HthFbPX7iUnT8uCKFUaT44sWgEtgM0ikgI0AX4SkQZYI4amDn2bAAdLaVeqbPyDYMwHEBhqTXifPkFUWCAzx3fno4mX0SoqhKmfbWfgtP+xMGk/efkF3o5YqUrJY8nCGPOzMaa+MSbGGBODlQi6GmMOA0uA222ronoBGcaYQ8ByYLCIRNgmtgfb2pQqu/BGMPoDyEiFRXdCgTWK6Na8LvMn9OKDO3sQGRrAI4u2MPjV1Xy+5SAFuoFPqULcuXQ2EVgHtBGRVBG5s5TuS4E9QDLwDnAvgDHmOPAssN72esbWplT5NOsJ106DX/8L/33G3iwi9I2N4pP7evPWbd3w8xEmzdvIdf9awzc7juiqKaVspDr+Y4iPjzdJSUneDkNVRp8/CEmzYORM6DCqyOH8AsOSzQd45avd7DueTXzzCCYPaUPPlvW8EKxSniUiG4wx8cUe02ShapS8s/De9XBwI3RKgK63Q+Nu1mS4g9z8AhYm7ef1/+4m7WQOfWMjmTykDR2b1PFS4Eq5nyYLpRxlpcPKKbBtMeRmQ1Rb6HKbVVcqtPBKujO5+Xyw7jemr0rmRHYuQ+Ma8NDg1sRGh3kpeKXcR5OFUsU5c9J6LOvGOZC6Hnz8oPVQa7TR6krw9bN3zTyTy8w1e5nx7V5Onc1jeOfGPDCoNc3qBXvxAyjlWposlHLmyA7Y+AFsng/ZRyGsIXQaa23qq9fK3u3EqbP853+/Mvu7FPILDAk9mvKngbFEhwd5MXilXEOThVJllZ8Lu5ZbiWP3CjAF0Oxy6HobtBtmPZkPSDt5hn99vZv5P+7H10cYf3kM91zRioiQAC9/AKUqTpOFUhVx8hBsTrRuUx3/FQJCof0I6HI7NIkHEfYdy+bVlbtYvOkAIQF+3NW3BXf2aUFYkL+3o1eq3DRZKHUxjIF931ujjXOT4pFtrFtUnRIgtD670jJ5ecUulm07TESwP/f2v4TbLmtOkL+vt6NXqsw0WSjlKjmZsPXcpPiP5yfFu9wGlwxi88Espq3Yybe7jxIdHsj9V8YyOr4p/r76nDFV+WmyUMod0neenxQ/lQ6h0bZJ8dtYlxHBtBU72fDbCZrVDebBq2K5oVNjfH2Kq42pVOWgyUIpd7JPis+xTYrnQ7PLMF1u5Vu/3jz/dSrbD52kdXQoDw1uw+B20Yho0lCVjyYLpTwl8/D5SfFjyRAQiokbztrwq3kqKZg9x7Lp1KQ2Dw9pQ59LIjVpqEpFk4VSnmafFJ9jmxQ/hanXmi31r+PJPR3YmhFIr5Z1mTykDd2a1/V2tEoBmiyU8q6cTCthbJwD+3/AiC/7I/vy6rGefJrdnv5tG/LQ4Da0axTu7UhVDafJQqnKIn2nlTQ2z4dTRzgVUI8FZ3szJ6cfcR3jeXBQLC2jQr0dpaqhNFkoVdnk51qT4RvnYHYtR0w+G0wbFuZfQUCHkdwzpBON69TydpSqhtFkoVRllpkGmxPJ2/ABfieSOWUCWWou4+SlY7nh2huJ0rpTqqzyzkLOSQiJrNDpmiyUqgqMgf0/cOqH2fj98gmBBafZYxqxv/kIulw/kfCoJt6OUHmTMZB9HDL2W48Izki94OtUyEqDZr3gjmUV+haaLJSqanKySP9+Pie/f5dWp7eShw/76/Wl0YC7CLz0avDV2lPVTl4OnDxw/gf/7/uLJoO804XP8asFtZs4vJpCVBuIu7FCIWiyUKoK2719A8nL3iI+YxlRksHpgHr4d7sZv663Q1Rrb4enysIYyD5W9Ie/4/ustKLnhUYXTgQXJobgekWe8ngxNFkoVQ1s2JvGf5fMpfPRzxnouxE/CjBNeiBdb4O44RCoT+/zmtwztlFBKbeI8s4UPsc/uOgPf8f34Y3BL9CjH0OThVLVhDGGNclHmfHlOtqkLeXWgNU0Mwcw/iFI3HCrEm6zXi79bbPGM8aq/VXaqOBU+gUnCYQ1KCUZNIVaEZXu70mThVLVjDGGFdvTeGn5DkLTN3J32HdcWbAWv7xTUO8SW/n0sdYPLFW63NPFjwTs7w9Afk7hc/xDoE4xt4XOfR3WCPyq3oOwNFkoVU3lFxiWbD7AK1/t5ujx49xb/2duD/qW8CNJIL4QO9h6yl/s4Jo5KV5QYBsVXJgMHL7OPlr4HPGxHqtb2i2ioDqVblTgCposlKrmzuYVsDBpP//6ejdpJ3O4KeY0k+snUX/Px9bEaUiU9dwNvyDrUbEY67/GOLwvrq2k98bJccdrluV65Y2Bsl0zNxvyzxb+wwoILX6y2D5X0KhmJlY0WShVY5zJzeeDdb8xfVUyJ7JzubpdJH9tc4DGez+C39ZancQHEOu/IsW8L66tpPdl6VPe69piLNN1HfoW18e/VtFkEFS7Wo4KXEGThVI1TOaZXGau2cuMb/dy6mwewzs3ZtLAS2gRGaJl0VWJNFkoVUMdP3WW//zvV977LoWcvAJq1/KnTXQYsdGhtGkQRuto61U3pOpNxirX02ShVA2XdvIMy7YeZmdaJrsOZ7IzLZPMM3n245GhgbRpEEps/TCHJBJKWFDNvHdfU5WWLPw8HYxSyvOiw4MYd3mM/b0xhrSTOfbksSvNei1Yv5/Tufn2fo1qB9HaYQTSJjqMS+qHUivA1wufQnmTJgulaiARoUHtIBrUDuKK1lH29oICw4HfT7PTNvrYnZbJzrQsvks+xtn8Atu50KxusH300TraGo20jAwlwM/HWx9JuZnbkoWIzAKuA44YY9rb2p4FhgEFwBFgvDHmoFgzbq8B1wDZtvafbOeMA/5qu+zfjDHvuStmpWo6Hx+had1gmtYNZlC7aHt7Xn4Bvx3Ptt/C2p2Wxc60TL7ecYT8AutWtq+P0CIyhDbR529jtW4QRvO6wfj5ahKp6tw2ZyEi/YAs4H2HZBFujDlp+/p+oJ0x5h4RuQb4E1ay6Am8ZozpKSJ1gSQgHmtl9QagmzHmRGnfW+cslPKMnLx89h49xU77rawsdqVlsu94Nud+tAT4+dAqKpQ2tuTR2jYv0rhOLXx8dGVWZeKVOQtjzGoRibmg7aTD2xCsBADWaON9Y2Wu70Wkjog0BPoDXxljjgOIyFfAUCDRXXErpcou0M+Xtg3Cadug8PPDT5/NJ/mINfo4Nx/y497jfLLpoL1PcIAvsfXP38aKtc2JRIcH6vLeSsjjcxYi8hxwO5ABDLA1Nwb2O3RLtbWV1F7cdScAEwCaNWvm2qCVUuVSK8CXDk1q06FJ7ULtJ8/ksts2+th5OJPdRzL5Zmc6H25ItfcJD/KzbmM1CDu/zDc6jHqhnq3AqgrzeLIwxjwJPCkijwOTgClY+zCLdC2lvbjrvg28DdZtKNdEq5RypfAgf7o1j6Bb84hC7cdPnbWPQHYetuZEvthyiHmn99n7RIYGFFnaGxsdRu1aurzXE7y5Gmoe8AVWskgFmjocawIctLX3v6B9lWfCU0p5St2QAHq1rEevlvXsbcYYjmTm2BPIuTmRD5P2c+rs+eW9DWsH2W5hhdqX+MZGhxIcoIs9Xcmjf5oiEmuM2W17ewOww/b1EmCSiMzHmuDOMMYcEpHlwN9F5NyvIYOBxz0Zs1LKO0SE6PAgosOD6BtbdHnv7iOZ7DycZR+RvLfnGGfzzi/vbRoRTKM6QYjtBoVxuCnhuK7HFPmiDH2xklnx7cX3p0z9S/i+pvTv76hNgzBeS+hS7LGL4c6ls4lYo4JIEUnFGkFcIyJtsJbO/gbcY+u+FGslVDLW0tk/ABhjjtuW26639Xvm3GS3UqpmclzeO7Dt+eW9+QWG346dsq/I2pWWSdrJM4XvZUvRL+0LsgR7YgEpVGuw0NcOFylpHt5xgl4KtRcbSpn6nztS8jWs/zasXav4oC6SlvtQSikFlL50VnfKKKWUckqThVJKKac0WSillHJKk4VSSimnNFkopZRySpOFUkoppzRZKKWUckqThVJKKaeq5aY8EUnH2iFeUZHAUReFU1XUtM9c0z4v6GeuKS7mMzc3xkQVd6BaJouLJSJJJe1irK5q2meuaZ8X9DPXFO76zHobSimllFOaLJRSSjmlyaJ4b3s7AC+oaZ+5pn1e0M9cU7jlM+uchVJKKad0ZKGUUsopTRZKKaWc0mRRDBGZKiIHRGST7XWNt2PyFBF5WESMiER6OxZ3E5FnRWSL7e94hYg08nZM7iYiL4rIDtvnXiwidbwdk7uJyE0isk1ECkSkWi+jFZGhIrJTRJJF5DFXXluTRcleMcZ0tr2WejsYTxCRpsBVwD5vx+IhLxpjOhpjOgOfA095OyAP+Apob4zpCOyiZjzTfiswAljt7UDcSUR8gTeAq4F2wFgRaeeq62uyUI5eAR7hgufMV1fGmJMOb0OoAZ/bGLPCGJNne/s90MSb8XiCMeYXY8xOb8fhAT2AZGPMHmPMWWA+MMxVF9dkUbJJtqH6LBGJ8HYw7iYiNwAHjDGbvR2LJ4nIcyKyH7iFmjGycHQH8KW3g1Au0xjY7/A+1dbmEn6uulBVIyIrgQbFHHoSeBN4Fus3zWeBl7D+YVVpTj7zE8Bgz0bkfqV9ZmPMp8aYJ4EnReRxYBIwxaMBuoGzz2zr8ySQB8z1ZGzuUpbPXANIMW0uGy3X2GRhjBlUln4i8g7W/ewqr6TPLCIdgBbAZhEB69bETyLSwxhz2IMhulxZ/56BecAXVINk4ewzi8g44DrgSlNNNlqV4++5OksFmjq8bwIcdNXF9TZUMUSkocPb4VgTZNWWMeZnY0x9Y0yMMSYG63+6rlU9UTgjIrEOb28AdngrFk8RkaHAo8ANxphsb8ejXGo9ECsiLUQkAEgAlrjq4jV2ZOHECyLSGWsIlwLc7d1wlJs8LyJtgAKskvb3eDkeT/g3EAh8ZRtFfm+MqdafW0SGA/8CooAvRGSTMWaIl8NyOWNMnohMApYDvsAsY8w2V11fy30opZRySm9DKaWUckqThVJKKac0WSillHJKk4VSSimnNFkopZRySpOFUkoppzRZKFUMEcly8fUm2cpGFyr/LpbXbce2iEhXh2MNReRzh/c9RGS1rQT1DhGZISLBInKdiDztyniVupAmC6U8Yy0wCGvzn6OrgVjbawJWXbJz/gK8AyAi0cCHwKPGmDbApcAyIAyrTMkNIhLszg+gajZNFkqVwvab/4sislVEfhaRMbZ2HxGZbnuozucislRERpV0HWPMRmNMSjGHhgHvG8v3QB2HcjMjsRICwH3Ae8aYdbbrGWPMImNMmq2+0yqsek9KuYUmC6VKNwLoDHTCGhm8aPthPgKIAToAdwGXVfD6xZaVFpEWwAljTI6tvT2woZTrJAF9KxiDUk5pslCqdH2ARGNMvjEmDfgf0N3W/qExpsBWcPGbCl6/pLLSDYH0clznCFDtHwurvEeThVKlK+6HeWnt5VVSWenTQJBD+zagWynXCbKdo5RbaLJQqnSrgTEi4isiUUA/4EdgDTDSNncRDfSv4PWXALfb5kZ6ARnGmENYz8eOcej3b2CciPQ81yAit4rIuQf+tKaal9JX3qXJQqnSLQa2AJuBr4FHbLedPsIaFWwF3gJ+ADJKuoiI3C8iqVgjhy0iMsN2aCmwB0jGWvl0L4Ax5hTwq4hcYnufhvV8gmm2pbO/YM1RnHuO+ACsVVFKuYWWKFeqgkQk1BiTJSL1sEYbvV35wCjbcxi6GWP+6qRfNDDPGHOlq763UhfShx8pVXGfi0gdIAB41tVPFjTGLLYlImeaAQ+58nsrdSEdWSjlQiKyGOt55o4eNcYs90Y8SrmKJgullFJO6QS3UkoppzRZKKWUckqThVJKKac0WSillHLq/wH15xVoBF4xjgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal value of c for validation set: 0.01\n"
     ]
    }
   ],
   "source": [
    "from sklearn import svm\n",
    "\n",
    "# Reduce training set for less computation time\n",
    "shrinking_factor = 0.01\n",
    "X_train_small = X_train_[:int(shrinking_factor*X_train_.shape[0])]\n",
    "y_train_log_small = y_train_log_[:int(shrinking_factor*X_train_.shape[0])]\n",
    "\n",
    "# Rough search for a good estimate of C on the interval [1,e-6]\n",
    "C = 1/np.power(10,np.arange(6))\n",
    "training_MAE = []; test_MAE = []\n",
    "for c in C:\n",
    "    SVM = svm.SVR(C=c, kernel='linear', cache_size=1000)\n",
    "    SVM.fit(X_train_small, y_train_log_small)\n",
    "    y_train_pred_log = SVM.predict(X_train_small)\n",
    "    y_val_pred_log = SVM.predict(X_val_)\n",
    "    training_MAE.append(np.average(abs(np.exp(y_train_log_small) - np.exp(y_train_pred_log))))\n",
    "    test_MAE.append(np.average(abs(np.exp(y_val_log_) - np.exp(y_val_pred_log))))\n",
    "\n",
    "plt.plot(np.log10(C), training_MAE,label='training MAE')\n",
    "plt.plot(np.log10(C), test_MAE,label='validation MAE')\n",
    "plt.xlabel('log_10(C)')\n",
    "plt.ylabel('MAE')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "\n",
    "cOpt = C[np.argmin(test_MAE)]\n",
    "print(\"Optimal value of c for validation set:\", cOpt)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training MAE using log transformation: 1296.36\n",
      "Test MAE using log transformation: 1308.77\n"
     ]
    }
   ],
   "source": [
    "# Evaluate the model on the test set\n",
    "\n",
    "# Cashed results\n",
    "MAE_train = 1296.36\n",
    "MAE_test = 1308.77\n",
    "\n",
    "modelAll = False # make true if you can wait ~ 30 mins\n",
    "if modelAll:\n",
    "    SVM = svm.SVR(C=cOpt, kernel='linear', cache_size=1000)\n",
    "    SVM.fit(X_train, y_train_log) \n",
    "    y_train_pred_log = SVM.predict(X_train)\n",
    "    y_test_pred_log = SVM.predict(X_test)\n",
    "    MAE_train = np.average(abs(np.exp(y_train_log) - np.exp(y_train_pred_log)))\n",
    "    MAE_test = np.average(abs(np.exp(y_test_log) - np.exp(y_test_pred_log)))\n",
    "    \n",
    "print(\"Training MAE using log transformation:\", MAE_train)\n",
    "print(\"Test MAE using log transformation:\", MAE_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Random Forest\n",
    "\n",
    "#### Fitting the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training MAE: 466.5807127869157\n",
      "Test MAE: 1251.667602352379\n"
     ]
    }
   ],
   "source": [
    "from sklearn.ensemble import RandomForestRegressor\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "\n",
    "# Gridsearch and model\n",
    "gridsearch = False\n",
    "if gridsearch:\n",
    "    parameters = {\n",
    "        'max_depth':range(50, 500, 50),\n",
    "        'max_features':range(10, 20, 2)\n",
    "    }\n",
    "    rf = RandomForestRegressor(n_estimators=50, oob_score=False)\n",
    "    grid = GridSearchCV(rf, parameters)\n",
    "    grid.fit(X_train, y_train)\n",
    "    print(\"Optimal parameters for the grid are %s with a score of %0.2f\"\n",
    "      % (grid.best_params_, grid.best_score_))\n",
    "\n",
    "    # Use our best RF model \n",
    "    rf = RandomForestRegressor(\n",
    "        n_estimators=50,\n",
    "        max_depth=grid.best_params_['max_depth'],\n",
    "        max_features=grid.best_params_['max_features'],\n",
    "        oob_score=False\n",
    "    )\n",
    "else:\n",
    "    # used cashed results from gridsearch\n",
    "    rf = RandomForestRegressor(n_estimators=50, max_depth=250, max_features=15)\n",
    "\n",
    "# Fit and test it\n",
    "rf.fit(X_train, y_train)\n",
    "y_train_pred = rf.predict(X_train)\n",
    "y_test_pred = rf.predict(X_test)\n",
    "print(\"Training MAE:\",np.average(abs(y_train - y_train_pred)))\n",
    "print(\"Test MAE:\",np.average(abs(y_test - y_test_pred)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Feature analysis\n",
    "\n",
    "RF did a good job modelling this problem. Tree methods are very useful for feature exploration so let's have a look. Looking at the top 20 features that best explain the output."
   ]
  },
  {
   "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"
    }
   ],
   "source": [
    "# Explore which features are the most important in RF model\n",
    "features = list(zip(X_train.columns, rf.feature_importances_))\n",
    "n = 20\n",
    "topFeatures = dict(sorted(features, key = lambda x: x[1], reverse=True)[1:n])\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "ax.barh(list(topFeatures.keys()), list(topFeatures.values()), color='green', ecolor='black')\n",
    "ax.set_yticks(list(topFeatures.keys()))\n",
    "ax.invert_yaxis()\n",
    "ax.set_xlabel('Importance')\n",
    "ax.set_ylabel('Feature')\n",
    "ax.set_title(\"Top {} features correlated with the output variable\".format(n))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Neural Network\n",
    "\n",
    "The last model is a simple neural network."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Re-loading data\n",
    "trainData = pd.read_csv(\"train.csv\")\n",
    "submissionData = pd.read_csv(\"test.csv\")\n",
    "\n",
    "# dropping correlated variables\n",
    "trainData.drop('cont12',axis=1,inplace=True)\n",
    "submissionData.drop('cont12',axis=1,inplace=True)\n",
    "\n",
    "# drop ID but store in submission data\n",
    "trainData.drop('id',axis=1,inplace=True)\n",
    "NN_y_train = trainData['loss']\n",
    "trainData.drop('loss',axis=1,inplace=True)\n",
    "submissionDataIDs = submissionData['id']\n",
    "submissionData.drop('id',axis=1,inplace=True)\n",
    "\n",
    "# one-hot-encoding \n",
    "NN_train_Data = pd.get_dummies(trainData)\n",
    "NN_sub_Data = pd.get_dummies(submissionData)\n",
    "trainLabels = list(NN_train_Data)\n",
    "subLabels = list(NN_sub_Data)\n",
    "\n",
    "for i in range(len(trainLabels)):\n",
    "    if (not (trainLabels[i] in subLabels)):\n",
    "        NN_train_Data.drop([trainLabels[i]],axis=1,inplace=True)\n",
    "        \n",
    "for i in range(len(subLabels)):\n",
    "    if (not (subLabels[i] in trainLabels)):\n",
    "        NN_sub_Data.drop([subLabels[i]],axis=1,inplace=True)\n",
    "\n",
    "# splitting into training and test set 80/20\n",
    "OneHotEncoded_X_train, OneHotEncoded_X_test, OneHotEncoded_y_train, OneHotEncoded_y_test = \\\n",
    "train_test_split(NN_train_Data, NN_y_train, test_size=0.2, random_state=20)\n",
    "\n",
    "# Feed it in as np.arrays\n",
    "OneHotEncoded_X_train = np.asarray(OneHotEncoded_X_train)\n",
    "OneHotEncoded_y_train = np.asarray(OneHotEncoded_y_train)\n",
    "OneHotEncoded_X_test = np.asarray(OneHotEncoded_X_test) \n",
    "NN_sub_Data = np.asarray(NN_sub_Data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 120523 samples, validate on 30131 samples\n",
      "Epoch 1/100\n",
      "120523/120523 [==============================] - 7s 55us/sample - loss: 1515.6954 - val_loss: 1283.3407\n",
      "Epoch 2/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1256.6484 - val_loss: 1221.8640\n",
      "Epoch 3/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1227.3377 - val_loss: 1213.2148\n",
      "Epoch 4/100\n",
      "120523/120523 [==============================] - 6s 49us/sample - loss: 1215.1058 - val_loss: 1204.2415\n",
      "Epoch 5/100\n",
      "120523/120523 [==============================] - 6s 49us/sample - loss: 1209.1157 - val_loss: 1191.8160\n",
      "Epoch 6/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1203.8813 - val_loss: 1188.4940\n",
      "Epoch 7/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1199.7916 - val_loss: 1198.0066\n",
      "Epoch 8/100\n",
      "120523/120523 [==============================] - 7s 56us/sample - loss: 1196.9303 - val_loss: 1243.9541\n",
      "Epoch 9/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1193.8551 - val_loss: 1195.1233\n",
      "Epoch 10/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1192.0581 - val_loss: 1194.7470\n",
      "Epoch 11/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1189.6686 - val_loss: 1186.2985\n",
      "Epoch 12/100\n",
      "120523/120523 [==============================] - 6s 49us/sample - loss: 1187.0830 - val_loss: 1179.8948\n",
      "Epoch 13/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1185.3740 - val_loss: 1178.2539\n",
      "Epoch 14/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1183.2043 - val_loss: 1181.7745\n",
      "Epoch 15/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1182.2053 - val_loss: 1176.1262\n",
      "Epoch 16/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1179.2628 - val_loss: 1171.2186\n",
      "Epoch 17/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1177.1201 - val_loss: 1178.4767\n",
      "Epoch 18/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1174.7778 - val_loss: 1198.6077\n",
      "Epoch 19/100\n",
      "120523/120523 [==============================] - 6s 53us/sample - loss: 1173.1508 - val_loss: 1166.3578\n",
      "Epoch 20/100\n",
      "120523/120523 [==============================] - 6s 54us/sample - loss: 1171.2971 - val_loss: 1166.3862\n",
      "Epoch 21/100\n",
      "120523/120523 [==============================] - 6s 54us/sample - loss: 1169.3730 - val_loss: 1171.2925\n",
      "Epoch 22/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1167.3546 - val_loss: 1166.3623\n",
      "Epoch 23/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1166.0555 - val_loss: 1197.9953\n",
      "Epoch 24/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1163.7576 - val_loss: 1163.8297\n",
      "Epoch 25/100\n",
      "120523/120523 [==============================] - 6s 53us/sample - loss: 1161.8399 - val_loss: 1165.8781\n",
      "Epoch 26/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1160.3590 - val_loss: 1159.7007\n",
      "Epoch 27/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1158.1793 - val_loss: 1164.4886\n",
      "Epoch 28/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1157.2968 - val_loss: 1164.7883\n",
      "Epoch 29/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1155.5202 - val_loss: 1160.7498\n",
      "Epoch 30/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1154.2244 - val_loss: 1158.5122\n",
      "Epoch 31/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1152.5087 - val_loss: 1157.0114\n",
      "Epoch 32/100\n",
      "120523/120523 [==============================] - 6s 53us/sample - loss: 1151.2989 - val_loss: 1154.7696\n",
      "Epoch 33/100\n",
      "120523/120523 [==============================] - 7s 55us/sample - loss: 1149.7208 - val_loss: 1162.0076\n",
      "Epoch 34/100\n",
      "120523/120523 [==============================] - 6s 53us/sample - loss: 1148.2140 - val_loss: 1158.5958\n",
      "Epoch 35/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1146.6173 - val_loss: 1154.2397\n",
      "Epoch 36/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1146.2032 - val_loss: 1152.7812\n",
      "Epoch 37/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1144.0562 - val_loss: 1153.2938\n",
      "Epoch 38/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1142.7006 - val_loss: 1149.5561\n",
      "Epoch 39/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1141.5109 - val_loss: 1150.3110\n",
      "Epoch 40/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1140.0783 - val_loss: 1166.9687\n",
      "Epoch 41/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1138.5944 - val_loss: 1169.3759\n",
      "Epoch 42/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1137.6716 - val_loss: 1151.0625\n",
      "Epoch 43/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1136.3139 - val_loss: 1149.4967\n",
      "Epoch 44/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1135.1829 - val_loss: 1171.0388\n",
      "Epoch 45/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1133.2831 - val_loss: 1149.1437\n",
      "Epoch 46/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1132.8095 - val_loss: 1147.6869\n",
      "Epoch 47/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1131.7998 - val_loss: 1153.5422\n",
      "Epoch 48/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1129.1766 - val_loss: 1147.9904\n",
      "Epoch 49/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1128.7547 - val_loss: 1158.5372\n",
      "Epoch 50/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1127.6484 - val_loss: 1154.7603\n",
      "Epoch 51/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1126.9015 - val_loss: 1143.4397\n",
      "Epoch 52/100\n",
      "120523/120523 [==============================] - 6s 53us/sample - loss: 1125.3876 - val_loss: 1167.4596\n",
      "Epoch 53/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1124.0938 - val_loss: 1147.5461\n",
      "Epoch 54/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1121.6989 - val_loss: 1149.5005\n",
      "Epoch 55/100\n",
      "120523/120523 [==============================] - 7s 54us/sample - loss: 1121.4300 - val_loss: 1147.0133\n",
      "Epoch 56/100\n",
      "120523/120523 [==============================] - 6s 53us/sample - loss: 1120.4945 - val_loss: 1145.4802\n",
      "Epoch 57/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1118.2556 - val_loss: 1144.7038\n",
      "Epoch 58/100\n",
      "120523/120523 [==============================] - 6s 53us/sample - loss: 1117.8577 - val_loss: 1142.3071\n",
      "Epoch 59/100\n",
      "120523/120523 [==============================] - 6s 53us/sample - loss: 1117.0639 - val_loss: 1156.6823\n",
      "Epoch 60/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1116.6381 - val_loss: 1148.5321\n",
      "Epoch 61/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1115.3524 - val_loss: 1143.2210\n",
      "Epoch 62/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1114.2902 - val_loss: 1142.7991\n",
      "Epoch 63/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1112.4634 - val_loss: 1148.3543\n",
      "Epoch 64/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1111.3056 - val_loss: 1142.5856\n",
      "Epoch 65/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1110.1622 - val_loss: 1143.5987\n",
      "Epoch 66/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1109.5330 - val_loss: 1155.4911\n",
      "Epoch 67/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1107.5427 - val_loss: 1146.6568\n",
      "Epoch 68/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1106.3334 - val_loss: 1161.1349\n",
      "Epoch 69/100\n",
      "120523/120523 [==============================] - 7s 55us/sample - loss: 1106.4429 - val_loss: 1142.7573\n",
      "Epoch 70/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1104.8975 - val_loss: 1154.4517\n",
      "Epoch 71/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1103.8874 - val_loss: 1142.9402\n",
      "Epoch 72/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1102.9304 - val_loss: 1150.0360\n",
      "Epoch 73/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1101.8913 - val_loss: 1169.6263\n",
      "Epoch 74/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1099.5974 - val_loss: 1142.0736\n",
      "Epoch 75/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1099.0520 - val_loss: 1177.8771\n",
      "Epoch 76/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1098.4644 - val_loss: 1144.6030\n",
      "Epoch 77/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1097.0715 - val_loss: 1149.8754\n",
      "Epoch 78/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1096.1138 - val_loss: 1163.7642\n",
      "Epoch 79/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1095.6049 - val_loss: 1141.0541\n",
      "Epoch 80/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1094.1125 - val_loss: 1144.7709\n",
      "Epoch 81/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1092.5312 - val_loss: 1139.2733\n",
      "Epoch 82/100\n",
      "120523/120523 [==============================] - 6s 50us/sample - loss: 1091.4227 - val_loss: 1162.4254\n",
      "Epoch 83/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1090.2200 - val_loss: 1141.2943\n",
      "Epoch 84/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1089.8543 - val_loss: 1140.7593\n",
      "Epoch 85/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1088.4602 - val_loss: 1144.8516\n",
      "Epoch 86/100\n",
      "120523/120523 [==============================] - 6s 54us/sample - loss: 1087.4472 - val_loss: 1143.4061\n",
      "Epoch 87/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1086.1080 - val_loss: 1149.3274\n",
      "Epoch 88/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1085.1669 - val_loss: 1146.2641\n",
      "Epoch 89/100\n",
      "120523/120523 [==============================] - 7s 55us/sample - loss: 1084.7319 - val_loss: 1150.0877\n",
      "Epoch 90/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1082.5354 - val_loss: 1152.8438\n",
      "Epoch 91/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1081.9470 - val_loss: 1143.5710\n",
      "Epoch 92/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1081.2162 - val_loss: 1178.7700\n",
      "Epoch 93/100\n",
      "120523/120523 [==============================] - 6s 51us/sample - loss: 1079.4466 - val_loss: 1146.0475\n",
      "Epoch 94/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1080.2471 - val_loss: 1147.6084\n",
      "Epoch 95/100\n",
      "120523/120523 [==============================] - 7s 56us/sample - loss: 1078.5550 - val_loss: 1163.1629\n",
      "Epoch 96/100\n",
      "120523/120523 [==============================] - 6s 53us/sample - loss: 1076.6158 - val_loss: 1146.6946\n",
      "Epoch 97/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1075.3836 - val_loss: 1142.9793\n",
      "Epoch 98/100\n",
      "120523/120523 [==============================] - 6s 53us/sample - loss: 1073.6881 - val_loss: 1143.6049\n",
      "Epoch 99/100\n",
      "120523/120523 [==============================] - 6s 52us/sample - loss: 1072.6841 - val_loss: 1173.0037\n",
      "Epoch 100/100\n",
      "120523/120523 [==============================] - 6s 53us/sample - loss: 1073.7768 - val_loss: 1163.0326\n"
     ]
    },
    {
     "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": [
      "Test MAE: 1179.8226424160018\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense, Activation, Dropout\n",
    "from tensorflow.keras.optimizers import SGD\n",
    "from sklearn.metrics import mean_absolute_error\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Dense(500, input_dim=OneHotEncoded_X_train.shape[1]))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dense(1))\n",
    "model.compile(loss='mae', optimizer='sgd')\n",
    "\n",
    "history = model.fit(OneHotEncoded_X_train, OneHotEncoded_y_train,\n",
    "                    epochs=100, batch_size=128, validation_split=0.2, shuffle=True,verbose=1)\n",
    "\n",
    "# Visualizing the loss\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.ylabel('MAE')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'validation'])\n",
    "plt.show()\n",
    "\n",
    "y_pred = model.predict(OneHotEncoded_X_test)\n",
    "test_error = mean_absolute_error(OneHotEncoded_y_test, y_pred)\n",
    "print(\"Test MAE:\", test_error)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Model the submission set (final test) with NN model (Best performance)\n",
    "y_pred_submission = model.predict(NN_sub_Data)\n",
    "with open(\"testSubmissionFile.csv\", \"w\") as subfile:\n",
    "    subfile.write(\"id,loss\\n\")\n",
    "    for i, prediction in enumerate(list(y_pred_submission)):\n",
    "        subfile.write(\"%s,%s\\n\"%(submissionDataIDs[i],prediction[0]))"
   ]
  }
 ],
 "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": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}