t81_558_class_04_3_regression.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# T81-558: Applications of Deep Neural Networks\n",
    "**Module 4: Training for Tabular Data**\n",
    "* Instructor: [Jeff Heaton](https://sites.wustl.edu/jeffheaton/), McKelvey School of Engineering, [Washington University in St. Louis](https://engineering.wustl.edu/Programs/Pages/default.aspx)\n",
    "* For more information visit the [class website](https://sites.wustl.edu/jeffheaton/t81-558/)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Module 4 Material\n",
    "\n",
    "* Part 4.1: Encoding a Feature Vector for Keras Deep Learning [[Video]](https://www.youtube.com/watch?v=Vxz-gfs9nMQ&list=PLjy4p-07OYzulelvJ5KVaT2pDlxivl_BN) [[Notebook]](t81_558_class_04_1_feature_encode.ipynb)\n",
    "* Part 4.2: Keras Multiclass Classification for Deep Neural Networks with ROC and AUC [[Video]](https://www.youtube.com/watch?v=-f3bg9dLMks&list=PLjy4p-07OYzulelvJ5KVaT2pDlxivl_BN) [[Notebook]](t81_558_class_04_2_multi_class.ipynb)\n",
    "* **Part 4.3: Keras Regression for Deep Neural Networks with RMSE** [[Video]](https://www.youtube.com/watch?v=wNhBUC6X5-E&list=PLjy4p-07OYzulelvJ5KVaT2pDlxivl_BN) [[Notebook]](t81_558_class_04_3_regression.ipynb)\n",
    "* Part 4.4: Backpropagation, Nesterov Momentum, and ADAM Neural Network Training [[Video]](https://www.youtube.com/watch?v=VbDg8aBgpck&list=PLjy4p-07OYzulelvJ5KVaT2pDlxivl_BN) [[Notebook]](t81_558_class_04_4_backprop.ipynb)\n",
    "* Part 4.5: Neural Network RMSE and Log Loss Error Calculation from Scratch [[Video]](https://www.youtube.com/watch?v=wmQX1t2PHJc&list=PLjy4p-07OYzulelvJ5KVaT2pDlxivl_BN) [[Notebook]](t81_558_class_04_5_rmse_logloss.ipynb)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Part 4.3: Keras Regression for Deep Neural Networks with RMSE\n",
    "\n",
    "Regression results are evaluated differently than classification.  Consider the following code that trains a neural network for regression on the data set **jh-simple-dataset.csv**.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from scipy.stats import zscore\n",
    "from sklearn.model_selection import train_test_split\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Read the data set\n",
    "df = pd.read_csv(\n",
    "    \"https://data.heatonresearch.com/data/t81-558/jh-simple-dataset.csv\",\n",
    "    na_values=['NA','?'])\n",
    "\n",
    "# Generate dummies for job\n",
    "df = pd.concat([df,pd.get_dummies(df['job'],prefix=\"job\")],axis=1)\n",
    "df.drop('job', axis=1, inplace=True)\n",
    "\n",
    "# Generate dummies for area\n",
    "df = pd.concat([df,pd.get_dummies(df['area'],prefix=\"area\")],axis=1)\n",
    "df.drop('area', axis=1, inplace=True)\n",
    "\n",
    "# Generate dummies for product\n",
    "df = pd.concat([df,pd.get_dummies(df['product'],prefix=\"product\")],axis=1)\n",
    "df.drop('product', axis=1, inplace=True)\n",
    "\n",
    "# Missing values for income\n",
    "med = df['income'].median()\n",
    "df['income'] = df['income'].fillna(med)\n",
    "\n",
    "# Standardize ranges\n",
    "df['income'] = zscore(df['income'])\n",
    "df['aspect'] = zscore(df['aspect'])\n",
    "df['save_rate'] = zscore(df['save_rate'])\n",
    "df['subscriptions'] = zscore(df['subscriptions'])\n",
    "\n",
    "# Convert to numpy - Classification\n",
    "x_columns = df.columns.drop('age').drop('id')\n",
    "x = df[x_columns].values\n",
    "y = df['age'].values\n",
    "\n",
    "# Create train/test\n",
    "x_train, x_test, y_train, y_test = train_test_split(    \n",
    "    x, y, test_size=0.25, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 1500 samples, validate on 500 samples\n",
      "Epoch 1/1000\n",
      "1500/1500 - 0s - loss: 2050.2592 - val_loss: 1688.9015\n",
      "Epoch 2/1000\n",
      "1500/1500 - 0s - loss: 1259.2252 - val_loss: 766.7950\n",
      "Epoch 3/1000\n",
      "1500/1500 - 0s - loss: 462.8568 - val_loss: 263.3388\n",
      "Epoch 4/1000\n",
      "1500/1500 - 0s - loss: 238.8222 - val_loss: 214.8102\n",
      "Epoch 5/1000\n",
      "1500/1500 - 0s - loss: 199.1456 - val_loss: 177.3204\n",
      "Epoch 6/1000\n",
      "1500/1500 - 0s - loss: 160.8790 - val_loss: 141.0083\n",
      "Epoch 7/1000\n",
      "1500/1500 - 0s - loss: 126.2230 - val_loss: 111.9455\n",
      "Epoch 8/1000\n",
      "1500/1500 - 0s - loss: 98.5313 - val_loss: 89.0690\n",
      "Epoch 9/1000\n",
      "1500/1500 - 0s - loss: 78.0905 - val_loss: 72.9670\n",
      "Epoch 10/1000\n",
      "1500/1500 - 0s - loss: 63.8341 - val_loss: 61.6533\n",
      "Epoch 11/1000\n",
      "1500/1500 - 0s - loss: 53.1875 - val_loss: 53.3525\n",
      "Epoch 12/1000\n",
      "1500/1500 - 0s - loss: 45.4890 - val_loss: 46.9061\n",
      "Epoch 13/1000\n",
      "1500/1500 - 0s - loss: 39.6459 - val_loss: 41.2554\n",
      "Epoch 14/1000\n",
      "1500/1500 - 0s - loss: 34.5316 - val_loss: 36.6225\n",
      "Epoch 15/1000\n",
      "1500/1500 - 0s - loss: 30.5361 - val_loss: 32.6542\n",
      "Epoch 16/1000\n",
      "1500/1500 - 0s - loss: 26.8931 - val_loss: 29.4575\n",
      "Epoch 17/1000\n",
      "1500/1500 - 0s - loss: 24.1103 - val_loss: 26.2343\n",
      "Epoch 18/1000\n",
      "1500/1500 - 0s - loss: 21.5168 - val_loss: 23.6881\n",
      "Epoch 19/1000\n",
      "1500/1500 - 0s - loss: 19.5780 - val_loss: 21.6607\n",
      "Epoch 20/1000\n",
      "1500/1500 - 0s - loss: 17.8715 - val_loss: 19.7502\n",
      "Epoch 21/1000\n",
      "1500/1500 - 0s - loss: 16.3498 - val_loss: 18.2493\n",
      "Epoch 22/1000\n",
      "1500/1500 - 0s - loss: 15.1986 - val_loss: 16.8494\n",
      "Epoch 23/1000\n",
      "1500/1500 - 0s - loss: 14.1297 - val_loss: 16.0222\n",
      "Epoch 24/1000\n",
      "1500/1500 - 0s - loss: 13.3175 - val_loss: 14.7425\n",
      "Epoch 25/1000\n",
      "1500/1500 - 0s - loss: 12.5616 - val_loss: 14.1886\n",
      "Epoch 26/1000\n",
      "1500/1500 - 0s - loss: 11.9911 - val_loss: 13.2288\n",
      "Epoch 27/1000\n",
      "1500/1500 - 0s - loss: 11.3565 - val_loss: 12.7648\n",
      "Epoch 28/1000\n",
      "1500/1500 - 0s - loss: 11.0307 - val_loss: 12.0009\n",
      "Epoch 29/1000\n",
      "1500/1500 - 0s - loss: 10.6133 - val_loss: 11.4967\n",
      "Epoch 30/1000\n",
      "1500/1500 - 0s - loss: 10.1955 - val_loss: 11.0097\n",
      "Epoch 31/1000\n",
      "1500/1500 - 0s - loss: 9.6920 - val_loss: 10.8203\n",
      "Epoch 32/1000\n",
      "1500/1500 - 0s - loss: 9.4271 - val_loss: 10.0933\n",
      "Epoch 33/1000\n",
      "1500/1500 - 0s - loss: 9.1585 - val_loss: 10.6077\n",
      "Epoch 34/1000\n",
      "1500/1500 - 0s - loss: 8.7490 - val_loss: 9.3496\n",
      "Epoch 35/1000\n",
      "1500/1500 - 0s - loss: 8.4172 - val_loss: 8.8913\n",
      "Epoch 36/1000\n",
      "1500/1500 - 0s - loss: 8.0880 - val_loss: 8.5390\n",
      "Epoch 37/1000\n",
      "1500/1500 - 0s - loss: 7.8997 - val_loss: 8.2320\n",
      "Epoch 38/1000\n",
      "1500/1500 - 0s - loss: 7.4462 - val_loss: 7.8511\n",
      "Epoch 39/1000\n",
      "1500/1500 - 0s - loss: 7.2211 - val_loss: 8.1105\n",
      "Epoch 40/1000\n",
      "1500/1500 - 0s - loss: 6.8694 - val_loss: 7.3071\n",
      "Epoch 41/1000\n",
      "1500/1500 - 0s - loss: 6.5722 - val_loss: 7.1561\n",
      "Epoch 42/1000\n",
      "1500/1500 - 0s - loss: 6.3514 - val_loss: 6.8532\n",
      "Epoch 43/1000\n",
      "1500/1500 - 0s - loss: 6.1179 - val_loss: 6.4865\n",
      "Epoch 44/1000\n",
      "1500/1500 - 0s - loss: 5.8204 - val_loss: 6.2660\n",
      "Epoch 45/1000\n",
      "1500/1500 - 0s - loss: 5.6559 - val_loss: 6.0099\n",
      "Epoch 46/1000\n",
      "1500/1500 - 0s - loss: 5.4042 - val_loss: 5.8534\n",
      "Epoch 47/1000\n",
      "1500/1500 - 0s - loss: 5.1679 - val_loss: 5.9630\n",
      "Epoch 48/1000\n",
      "1500/1500 - 0s - loss: 4.9686 - val_loss: 5.2570\n",
      "Epoch 49/1000\n",
      "1500/1500 - 0s - loss: 4.6872 - val_loss: 4.9776\n",
      "Epoch 50/1000\n",
      "1500/1500 - 0s - loss: 4.3808 - val_loss: 4.7889\n",
      "Epoch 51/1000\n",
      "1500/1500 - 0s - loss: 4.1959 - val_loss: 4.5818\n",
      "Epoch 52/1000\n",
      "1500/1500 - 0s - loss: 3.9644 - val_loss: 4.3277\n",
      "Epoch 53/1000\n",
      "1500/1500 - 0s - loss: 3.7148 - val_loss: 4.1350\n",
      "Epoch 54/1000\n",
      "1500/1500 - 0s - loss: 3.5353 - val_loss: 4.2233\n",
      "Epoch 55/1000\n",
      "1500/1500 - 0s - loss: 3.4220 - val_loss: 3.7993\n",
      "Epoch 56/1000\n",
      "1500/1500 - 0s - loss: 3.2501 - val_loss: 3.5910\n",
      "Epoch 57/1000\n",
      "1500/1500 - 0s - loss: 3.0846 - val_loss: 3.4863\n",
      "Epoch 58/1000\n",
      "1500/1500 - 0s - loss: 2.8832 - val_loss: 3.2905\n",
      "Epoch 59/1000\n",
      "1500/1500 - 0s - loss: 2.7629 - val_loss: 3.4687\n",
      "Epoch 60/1000\n",
      "1500/1500 - 0s - loss: 2.6956 - val_loss: 3.0951\n",
      "Epoch 61/1000\n",
      "1500/1500 - 0s - loss: 2.5179 - val_loss: 3.2593\n",
      "Epoch 62/1000\n",
      "1500/1500 - 0s - loss: 2.4358 - val_loss: 2.7662\n",
      "Epoch 63/1000\n",
      "1500/1500 - 0s - loss: 2.3071 - val_loss: 2.6290\n",
      "Epoch 64/1000\n",
      "1500/1500 - 0s - loss: 2.1788 - val_loss: 2.5419\n",
      "Epoch 65/1000\n",
      "1500/1500 - 0s - loss: 2.0924 - val_loss: 2.4129\n",
      "Epoch 66/1000\n",
      "1500/1500 - 0s - loss: 1.9647 - val_loss: 2.3891\n",
      "Epoch 67/1000\n",
      "1500/1500 - 0s - loss: 1.8656 - val_loss: 2.1799\n",
      "Epoch 68/1000\n",
      "1500/1500 - 0s - loss: 1.7737 - val_loss: 2.1117\n",
      "Epoch 69/1000\n",
      "1500/1500 - 0s - loss: 1.6713 - val_loss: 1.9839\n",
      "Epoch 70/1000\n",
      "1500/1500 - 0s - loss: 1.6029 - val_loss: 1.9519\n",
      "Epoch 71/1000\n",
      "1500/1500 - 0s - loss: 1.5406 - val_loss: 1.9135\n",
      "Epoch 72/1000\n",
      "1500/1500 - 0s - loss: 1.4558 - val_loss: 1.7606\n",
      "Epoch 73/1000\n",
      "1500/1500 - 0s - loss: 1.3821 - val_loss: 1.6311\n",
      "Epoch 74/1000\n",
      "1500/1500 - 0s - loss: 1.3134 - val_loss: 1.5733\n",
      "Epoch 75/1000\n",
      "1500/1500 - 0s - loss: 1.2986 - val_loss: 1.6177\n",
      "Epoch 76/1000\n",
      "1500/1500 - 0s - loss: 1.1986 - val_loss: 1.4329\n",
      "Epoch 77/1000\n",
      "1500/1500 - 0s - loss: 1.1612 - val_loss: 1.4040\n",
      "Epoch 78/1000\n",
      "1500/1500 - 0s - loss: 1.0643 - val_loss: 1.5199\n",
      "Epoch 79/1000\n",
      "1500/1500 - 0s - loss: 1.0872 - val_loss: 1.3153\n",
      "Epoch 80/1000\n",
      "1500/1500 - 0s - loss: 1.0027 - val_loss: 1.5200\n",
      "Epoch 81/1000\n",
      "1500/1500 - 0s - loss: 1.0093 - val_loss: 1.1644\n",
      "Epoch 82/1000\n",
      "1500/1500 - 0s - loss: 0.9259 - val_loss: 1.1706\n",
      "Epoch 83/1000\n",
      "1500/1500 - 0s - loss: 0.9206 - val_loss: 1.0793\n",
      "Epoch 84/1000\n",
      "1500/1500 - 0s - loss: 0.8240 - val_loss: 1.0378\n",
      "Epoch 85/1000\n",
      "1500/1500 - 0s - loss: 0.7869 - val_loss: 0.9985\n",
      "Epoch 86/1000\n",
      "1500/1500 - 0s - loss: 0.7815 - val_loss: 0.9728\n",
      "Epoch 87/1000\n",
      "1500/1500 - 0s - loss: 0.7721 - val_loss: 0.9672\n",
      "Epoch 88/1000\n",
      "1500/1500 - 0s - loss: 0.7720 - val_loss: 0.9470\n",
      "Epoch 89/1000\n",
      "1500/1500 - 0s - loss: 0.6813 - val_loss: 0.9458\n",
      "Epoch 90/1000\n",
      "1500/1500 - 0s - loss: 0.6726 - val_loss: 0.8536\n",
      "Epoch 91/1000\n",
      "1500/1500 - 0s - loss: 0.6359 - val_loss: 0.8321\n",
      "Epoch 92/1000\n",
      "1500/1500 - 0s - loss: 0.6420 - val_loss: 0.9361\n",
      "Epoch 93/1000\n",
      "1500/1500 - 0s - loss: 0.6398 - val_loss: 0.7965\n",
      "Epoch 94/1000\n",
      "1500/1500 - 0s - loss: 0.6042 - val_loss: 0.7671\n",
      "Epoch 95/1000\n",
      "1500/1500 - 0s - loss: 0.5727 - val_loss: 0.7501\n",
      "Epoch 96/1000\n",
      "1500/1500 - 0s - loss: 0.5955 - val_loss: 0.7549\n",
      "Epoch 97/1000\n",
      "1500/1500 - 0s - loss: 0.5514 - val_loss: 0.7067\n",
      "Epoch 98/1000\n",
      "1500/1500 - 0s - loss: 0.5409 - val_loss: 0.6884\n",
      "Epoch 99/1000\n",
      "1500/1500 - 0s - loss: 0.5777 - val_loss: 0.8978\n",
      "Epoch 100/1000\n",
      "1500/1500 - 0s - loss: 0.5352 - val_loss: 0.6997\n",
      "Epoch 101/1000\n",
      "1500/1500 - 0s - loss: 0.5110 - val_loss: 0.7224\n",
      "Epoch 102/1000\n",
      "1500/1500 - 0s - loss: 0.5055 - val_loss: 0.7285\n",
      "Epoch 103/1000\n",
      "Restoring model weights from the end of the best epoch.\n",
      "1500/1500 - 0s - loss: 0.5336 - val_loss: 0.7553\n",
      "Epoch 00103: early stopping\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.keras.callbacks.History at 0x1a31eb3a58>"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense, Activation\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "\n",
    "# Build the neural network\n",
    "model = Sequential()\n",
    "model.add(Dense(25, input_dim=x.shape[1], activation='relu')) # Hidden 1\n",
    "model.add(Dense(10, activation='relu')) # Hidden 2\n",
    "model.add(Dense(1)) # Output\n",
    "model.compile(loss='mean_squared_error', optimizer='adam')\n",
    "monitor = EarlyStopping(monitor='val_loss', min_delta=1e-3, \n",
    "                        patience=5, verbose=1, mode='auto', restore_best_weights=True)\n",
    "model.fit(x_train,y_train,validation_data=(x_test,y_test),callbacks=[monitor],verbose=2,epochs=1000)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Mean Square Error\n",
    "\n",
    "The mean square error is the sum of the squared differences between the prediction ($\\hat{y}$) and the expected ($y$).  MSE values are not of a particular unit.  If an MSE value has decreased for a model, that is good.  However, beyond this, there is not much more you can determine.  Low MSE values are desired.\n",
    "\n",
    "$ \\text{MSE} = \\frac{1}{n} \\sum_{i=1}^n \\left(\\hat{y}_i - y_i\\right)^2 $\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Final score (MSE): 0.6883844942541036\n"
     ]
    }
   ],
   "source": [
    "from sklearn import metrics\n",
    "\n",
    "# Predict\n",
    "pred = model.predict(x_test)\n",
    "\n",
    "# Measure MSE error.  \n",
    "score = metrics.mean_squared_error(pred,y_test)\n",
    "print(\"Final score (MSE): {}\".format(score))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Root Mean Square Error\n",
    "\n",
    "The root mean square (RMSE) is essentially the square root of the MSE.  Because of this, the RMSE error is in the same units as the training data outcome. Low RMSE values are desired.\n",
    "\n",
    "$ \\text{RMSE} = \\sqrt{\\frac{1}{n} \\sum_{i=1}^n \\left(\\hat{y}_i - y_i\\right)^2} $"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Final score (RMSE): 0.8296893962526601\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "\n",
    "# Measure RMSE error.  RMSE is common for regression.\n",
    "score = np.sqrt(metrics.mean_squared_error(pred,y_test))\n",
    "print(\"Final score (RMSE): {}\".format(score))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Lift Chart\n",
    "\n",
    "\n",
    "To generate a lift chart, perform the following activities:\n",
    "\n",
    "* Sort the data by expected output. Plot the blue line above.\n",
    "* For every point on the x-axis plot the predicted value for that same data point. This is the green line above.\n",
    "* The x-axis is just 0 to 100% of the dataset. The expected always starts low and ends high.\n",
    "* The y-axis is ranged according to the values predicted.\n",
    "\n",
    "Reading a lift chart:\n",
    "\n",
    "* The expected and predict lines should be close. Notice where one is above the ot other.\n",
    "* The below chart is the most accurate on lower age."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Regression chart.\n",
    "def chart_regression(pred, y, sort=True):\n",
    "    t = pd.DataFrame({'pred': pred, 'y': y.flatten()})\n",
    "    if sort:\n",
    "        t.sort_values(by=['y'], inplace=True)\n",
    "    plt.plot(t['y'].tolist(), label='expected')\n",
    "    plt.plot(t['pred'].tolist(), label='prediction')\n",
    "    plt.ylabel('output')\n",
    "    plt.legend()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "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": [
    "\n",
    "# Plot the chart\n",
    "chart_regression(pred.flatten(),y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python 3.6 (tensorflow)",
   "language": "python",
   "name": "rga"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}