t81_558_class_08_3_keras_hyperparameters.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# T81-558: Applications of Deep Neural Networks\n",
    "**Module 8: Kaggle Data Sets**\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 8 Material\n",
    "\n",
    "* Part 8.1: Introduction to Kaggle [[Video]](https://www.youtube.com/watch?v=XpGI4engRjQ&list=PLjy4p-07OYzulelvJ5KVaT2pDlxivl_BN&index=24) [[Notebook]](t81_558_class_08_1_kaggle_intro.ipynb)\n",
    "* Part 8.2: Building Ensembles with Scikit-Learn and Keras [[Video]](https://www.youtube.com/watch?v=AA3KFxjPxCo&index=25&list=PLjy4p-07OYzulelvJ5KVaT2pDlxivl_BN) [[Notebook]](t81_558_class_08_2_keras_ensembles.ipynb)\n",
    "* **Part 8.3: How Should you Architect Your Keras Neural Network: Hyperparameters** [[Video]](https://www.youtube.com/watch?v=GaKo-9c532c) [[Notebook]](t81_558_class_08_3_keras_hyperparameters.ipynb)\n",
    "* Part 8.4: Bayesian Hyperparameter Optimization for Keras [[Video]](https://www.youtube.com/watch?v=GaKo-9c532c) [[Notebook]](t81_558_class_08_4_bayesian_hyperparameter_opt.ipynb)\n",
    "* Part 8.5: Current Semester's Kaggle [[Video]](https://www.youtube.com/watch?v=GaKo-9c532c) [[Notebook]](t81_558_class_08_5_kaggle_project.ipynb)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Nicely formatted time string\n",
    "def hms_string(sec_elapsed):\n",
    "    h = int(sec_elapsed / (60 * 60))\n",
    "    m = int((sec_elapsed % (60 * 60)) / 60)\n",
    "    s = sec_elapsed % 60\n",
    "    return \"{}:{:>02}:{:>05.2f}\".format(h, m, s)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Part 8.3: How Should you Architect Your Keras Neural Network: Hyperparameters\n",
    "\n",
    "* [Guide to choosing Hyperparameters for your Neural Networks](https://towardsdatascience.com/guide-to-choosing-hyperparameters-for-your-neural-networks-38244e87dafe)\n",
    "\n",
    "### Number of Hidden Layers and Neuron Counts\n",
    "\n",
    "* [Keras Layers](https://keras.io/layers/core/)\n",
    "\n",
    "Layer types and when to use them:\n",
    "\n",
    "* **Activation** - Layer that simply adds an activation function, the activation function can also be specified as part of a Dense (or other) layer type.\n",
    "* **ActivityRegularization** Used to add L1/L2 regularization outside of a layer.  L1 and L2 can also be specified as part of a Dense (or other) layer type.\n",
    "* **Dense** - The original neural network layer type.  Every neuron is connected to the next layer.  The input vector is one-dimensional and placing certain inputs next to each other does not have an effect. \n",
    "* **Dropout** - Dropout consists in randomly setting a fraction rate of input units to 0 at each update during training time, which helps prevent overfitting.  Dropout only occurs during training.\n",
    "* **Flatten** - Flattens the input to 1D. Does not affect the batch size.\n",
    "* **Input** - A Keras tensor is a tensor object from the underlying backend (Theano, TensorFlow or CNTK), which we augment with certain attributes that allow us to build a Keras model just by knowing the inputs and outputs of the model.\n",
    "* **Lambda** - Wraps arbitrary expression as a Layer object.\n",
    "* **Masking** - Masks a sequence by using a mask value to skip timesteps.\n",
    "* **Permute** - Permutes the dimensions of the input according to a given pattern. Useful for e.g. connecting RNNs and convnets together.\n",
    "* **RepeatVector** - Repeats the input n times.\n",
    "* **Reshape** - Similar to Numpy reshapes.\n",
    "* **SpatialDropout1D** - This version performs the same function as Dropout, however it drops entire 1D feature maps instead of individual elements. \n",
    "* **SpatialDropout2D** - This version performs the same function as Dropout, however it drops entire 2D feature maps instead of individual elements\n",
    "* **SpatialDropout3D** - This version performs the same function as Dropout, however it drops entire 3D feature maps instead of individual elements. \n",
    "\n",
    "\n",
    "### Activation Functions\n",
    "\n",
    "* [Keras Activation Functions](https://keras.io/activations/)\n",
    "* [Activation Function Cheat Sheets](https://ml-cheatsheet.readthedocs.io/en/latest/activation_functions.html)\n",
    "\n",
    "As follows:\n",
    "\n",
    "* **softmax** - Used for multi-class classification.  Ensures all output neurons behave as probabilities and sum to 1.0.\n",
    "* **elu** - Exponential linear unit.  Exponential Linear Unit or its widely known name ELU is a function that tend to converge cost to zero faster and produce more accurate results.  Can produce negative outputs.\n",
    "* **selu** - Scaled Exponential Linear Unit (SELU), essentially **elu** multiplied by a scaling constant.\n",
    "* **softplus** - Softplus activation function. $log(exp(x) + 1)$  [Introduced](https://papers.nips.cc/paper/1920-incorporating-second-order-functional-knowledge-for-better-option-pricing.pdf) in 2001.\n",
    "* **softsign** Softsign activation function. $x / (abs(x) + 1)$ Similar to tanh, but not widely used.\n",
    "* **relu** - Very popular neural network activation function.  Used for hidden layers, cannot output negative values.  No trainable parameters.\n",
    "* **tanh** Classic neural network activation function, though often replaced by relu family on modern networks.\n",
    "* **sigmoid** - Classic neural network activation.  Often used on output layer of a binary classifier.\n",
    "* **hard_sigmoid** - Less computationally expensive variant of sigmoid.\n",
    "* **exponential** - Exponential (base e) activation function.\n",
    "* **linear** - Pass through activation function. Usually used on the output layer of a regression neural network.\n",
    "\n",
    "### Advanced Activation Functions\n",
    "\n",
    "* [Keras Advanced Activation Functions](https://keras.io/layers/advanced-activations/)\n",
    "\n",
    "The advanced activation functions contain parameters that are trained during neural network fitting. As follows:\n",
    "\n",
    "* **LeakyReLU** - Leaky version of a Rectified Linear Unit. It allows a small gradient when the unit is not active, controlled by alpha hyperparameter.\n",
    "* **PReLU** - Parametric Rectified Linear Unit, learns the alpha hyperparameter. \n",
    "\n",
    "### Regularization: L1, L2, Dropout\n",
    "\n",
    "* [Keras Regularization](https://keras.io/regularizers/)\n",
    "* [Keras Dropout](https://keras.io/layers/core/)\n",
    "\n",
    "### Batch Normalization\n",
    "\n",
    "* [Keras Batch Normalization](https://keras.io/layers/normalization/)\n",
    "\n",
    "* Ioffe, S., & Szegedy, C. (2015). [Batch normalization: Accelerating deep network training by reducing internal covariate shift](https://arxiv.org/abs/1502.03167). *arXiv preprint arXiv:1502.03167*.\n",
    "\n",
    "Normalize the activations of the previous layer at each batch, i.e. applies a transformation that maintains the mean activation close to 0 and the activation standard deviation close to 1.  Can allow learning rate to be larger.\n",
    "\n",
    "\n",
    "### Training Parameters\n",
    "\n",
    "* [Keras Optimizers](https://keras.io/optimizers/)\n",
    "\n",
    "* **Batch Size** - Usually small, such as 32 or so.\n",
    "* **Learning Rate**  - Usually small, 1e-3 or so."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from scipy.stats import zscore\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",
    "# 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['age'] = zscore(df['age'])\n",
    "df['subscriptions'] = zscore(df['subscriptions'])\n",
    "\n",
    "# Convert to numpy - Classification\n",
    "x_columns = df.columns.drop('product').drop('id')\n",
    "x = df[x_columns].values\n",
    "dummies = pd.get_dummies(df['product']) # Classification\n",
    "products = dummies.columns\n",
    "y = dummies.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import os\n",
    "import numpy as np\n",
    "import time\n",
    "import tensorflow.keras.initializers\n",
    "import statistics\n",
    "import tensorflow.keras\n",
    "from sklearn import metrics\n",
    "from sklearn.model_selection import StratifiedKFold\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense, Activation, Dropout\n",
    "from tensorflow.keras import regularizers\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "from sklearn.model_selection import StratifiedShuffleSplit\n",
    "from tensorflow.keras.layers import LeakyReLU,PReLU\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "\n",
    "\n",
    "\n",
    "def evaluate_network(dropout,lr,neuronPct,neuronShrink):\n",
    "    SPLITS = 2\n",
    "\n",
    "    # Bootstrap\n",
    "    boot = StratifiedShuffleSplit(n_splits=SPLITS, test_size=0.1)\n",
    "\n",
    "    # Track progress\n",
    "    mean_benchmark = []\n",
    "    epochs_needed = []\n",
    "    num = 0\n",
    "    neuronCount = int(neuronPct * 5000)\n",
    "\n",
    "    # Loop through samples\n",
    "    for train, test in boot.split(x,df['product']):\n",
    "        start_time = time.time()\n",
    "        num+=1\n",
    "\n",
    "        # Split train and test\n",
    "        x_train = x[train]\n",
    "        y_train = y[train]\n",
    "        x_test = x[test]\n",
    "        y_test = y[test]\n",
    "\n",
    "        # Construct neural network\n",
    "        # kernel_initializer = tensorflow.keras.initializers.he_uniform(seed=None)\n",
    "        model = Sequential()\n",
    "        \n",
    "        layer = 0\n",
    "        while neuronCount>25 and layer<10:\n",
    "            #print(neuronCount)\n",
    "            if layer==0:\n",
    "                model.add(Dense(neuronCount, \n",
    "                    input_dim=x.shape[1], \n",
    "                    activation=PReLU())) \n",
    "            else:\n",
    "                model.add(Dense(neuronCount, activation=PReLU())) \n",
    "            model.add(Dropout(dropout))\n",
    "        \n",
    "            neuronCount = neuronCount * neuronShrink\n",
    "        \n",
    "        model.add(Dense(y.shape[1],activation='softmax')) # Output\n",
    "        model.compile(loss='categorical_crossentropy', optimizer=Adam(lr=lr))\n",
    "        monitor = EarlyStopping(monitor='val_loss', min_delta=1e-3, \n",
    "            patience=100, verbose=0, mode='auto', restore_best_weights=True)\n",
    "\n",
    "        # Train on the bootstrap sample\n",
    "        model.fit(x_train,y_train,validation_data=(x_test,y_test),callbacks=[monitor],verbose=0,epochs=1000)\n",
    "        epochs = monitor.stopped_epoch\n",
    "        epochs_needed.append(epochs)\n",
    "\n",
    "        # Predict on the out of boot (validation)\n",
    "        pred = model.predict(x_test)\n",
    "\n",
    "        # Measure this bootstrap's log loss\n",
    "        y_compare = np.argmax(y_test,axis=1) # For log loss calculation\n",
    "        score = metrics.log_loss(y_compare, pred)\n",
    "        mean_benchmark.append(score)\n",
    "        m1 = statistics.mean(mean_benchmark)\n",
    "        m2 = statistics.mean(epochs_needed)\n",
    "        mdev = statistics.pstdev(mean_benchmark)\n",
    "\n",
    "        # Record this iteration\n",
    "        time_took = time.time() - start_time\n",
    "        #print(f\"#{num}: score={score:.6f}, mean score={m1:.6f}, stdev={mdev:.6f}, epochs={epochs}, mean epochs={int(m2)}, time={hms_string(time_took)}\")\n",
    "\n",
    "    tensorflow.keras.backend.clear_session()\n",
    "    return (-m1)\n",
    "\n",
    "print(evaluate_network(\n",
    "    dropout=0.2,\n",
    "    lr=1e-3,\n",
    "    neuronPct=0.2,\n",
    "    neuronShrink=0.2))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Part 8.4: Bayesian Hyperparameter Optimization for Keras\n",
    "\n",
    "Snoek, J., Larochelle, H., & Adams, R. P. (2012). [Practical bayesian optimization of machine learning algorithms](https://arxiv.org/pdf/1206.2944.pdf). In *Advances in neural information processing systems* (pp. 2951-2959).\n",
    "\n",
    "\n",
    "* [bayesian-optimization](https://github.com/fmfn/BayesianOptimization)\n",
    "* [hyperopt](https://github.com/hyperopt/hyperopt)\n",
    "* [spearmint](https://github.com/JasperSnoek/spearmint)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!pip install bayesian-optimization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from bayes_opt import BayesianOptimization\n",
    "import time\n",
    "\n",
    "# Bounded region of parameter space\n",
    "pbounds = {'dropout': (0.0, 0.499),\n",
    "           'lr': (0.0, 0.1),\n",
    "           'neuronPct': (0.01, 1),\n",
    "           'neuronShrink': (0.01, 1)\n",
    "          }\n",
    "\n",
    "optimizer = BayesianOptimization(\n",
    "    f=evaluate_network,\n",
    "    pbounds=pbounds,\n",
    "    verbose=2,  # verbose = 1 prints only when a maximum is observed, verbose = 0 is silent\n",
    "    random_state=1,\n",
    ")\n",
    "\n",
    "start_time = time.time()\n",
    "optimizer.maximize(init_points=10, n_iter=100,)\n",
    "time_took = time.time() - start_time\n",
    "\n",
    "print(optimizer.max)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "{'target': -0.5995399297989206, 'params': {'dropout': 0.13361186083708476, 'lr': 0.014767159034158229, 'neuronPct': 0.01100802893634497, 'neuronShrink': 0.7109218810227248}}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Part 8.5: Current Semester's Kaggle\n",
    "\n",
    "Kaggke competition site for current semester (Fall 2019):\n",
    "\n",
    "* Coming soon\n",
    "\n",
    "Previous Kaggle competition sites for this class (NOT this semester's assignment, feel free to use code):\n",
    "* [Spring 2019 Kaggle Assignment](https://www.kaggle.com/c/applications-of-deep-learningwustl-spring-2019)\n",
    "* [Fall 2018 Kaggle Assignment](https://www.kaggle.com/c/wustl-t81-558-washu-deep-learning-fall-2018)\n",
    "* [Spring 2018 Kaggle Assignment](https://www.kaggle.com/c/wustl-t81-558-washu-deep-learning-spring-2018)\n",
    "* [Fall 2017 Kaggle Assignment](https://www.kaggle.com/c/wustl-t81-558-washu-deep-learning-fall-2017)\n",
    "* [Spring 2017 Kaggle Assignment](https://inclass.kaggle.com/c/applications-of-deep-learning-wustl-spring-2017)\n",
    "* [Fall 2016 Kaggle Assignment](https://inclass.kaggle.com/c/wustl-t81-558-washu-deep-learning-fall-2016)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Iris as a Kaggle Competition\n",
    "\n",
    "If the Iris data were used as a Kaggle, you would be given the following three files:\n",
    "\n",
    "* [kaggle_iris_test.csv](https://data.heatonresearch.com/data/t81-558/datasets/kaggle_iris_test.csv) - The data that Kaggle will evaluate you on.  Contains only input, you must provide answers.  (contains x)\n",
    "* [kaggle_iris_train.csv](https://data.heatonresearch.com/data/t81-558/datasets/kaggle_iris_train.csv) - The data that you will use to train. (contains x and y)\n",
    "* [kaggle_iris_sample.csv](https://data.heatonresearch.com/data/t81-558/datasets/kaggle_iris_sample.csv) - A sample submission for Kaggle. (contains x and y)\n",
    "\n",
    "Important features of the Kaggle iris files (that differ from how we've previously seen files):\n",
    "\n",
    "* The iris species is already index encoded.\n",
    "* Your training data is in a separate file.\n",
    "* You will load the test data to generate a submission file.\n",
    "\n",
    "The following program generates a submission file for \"Iris Kaggle\".  You can use it as a starting point for assignment 3."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of classes: 3\n",
      "Restoring model weights from the end of the best epoch.\n",
      "Epoch 00096: early stopping\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.keras.callbacks.History at 0x1a33005e10>"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import os\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense, Activation\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "\n",
    "df_train = pd.read_csv(\"https://data.heatonresearch.com/data/t81-558/datasets/kaggle_iris_train.csv\",\n",
    "                       na_values=['NA','?'])\n",
    "\n",
    "# Encode feature vector\n",
    "df_train.drop('id', axis=1, inplace=True)\n",
    "\n",
    "num_classes = len(df_train.groupby('species').species.nunique())\n",
    "\n",
    "print(\"Number of classes: {}\".format(num_classes))\n",
    "\n",
    "# Convert to numpy - Classification\n",
    "x = df_train[['sepal_l', 'sepal_w', 'petal_l', 'petal_w']].values\n",
    "dummies = pd.get_dummies(df_train['species']) # Classification\n",
    "species = dummies.columns\n",
    "y = dummies.values\n",
    "    \n",
    "# Split into train/test\n",
    "x_train, x_test, y_train, y_test = train_test_split(    \n",
    "    x, y, test_size=0.25, random_state=45)\n",
    "\n",
    "# Train, with early stopping\n",
    "model = Sequential()\n",
    "model.add(Dense(50, input_dim=x.shape[1], activation='relu'))\n",
    "model.add(Dense(25))\n",
    "model.add(Dense(y.shape[1],activation='softmax'))\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam')\n",
    "monitor = EarlyStopping(monitor='val_loss', min_delta=1e-3, patience=5, verbose=1, mode='auto',\n",
    "                       restore_best_weights=True)\n",
    "\n",
    "model.fit(x_train,y_train,validation_data=(x_test,y_test),callbacks=[monitor],verbose=0,epochs=1000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Log loss score: 0.13537994282320143\n"
     ]
    }
   ],
   "source": [
    "from sklearn import metrics\n",
    "\n",
    "# Calculate multi log loss error\n",
    "pred = model.predict(x_test)\n",
    "score = metrics.log_loss(y_test, pred)\n",
    "print(\"Log loss score: {}\".format(score))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "     id  species-0  species-1     species-2\n",
      "0   100   0.013904   0.795336  1.907604e-01\n",
      "1   101   0.001614   0.381189  6.171972e-01\n",
      "2   102   0.002317   0.466447  5.312361e-01\n",
      "3   103   0.973577   0.026410  1.352797e-05\n",
      "4   104   0.981406   0.018586  8.082319e-06\n",
      "5   105   0.979300   0.020693  6.637653e-06\n",
      "6   106   0.994228   0.005771  5.370086e-07\n",
      "7   107   0.000977   0.256403  7.426196e-01\n",
      "8   108   0.017728   0.843886  1.383867e-01\n",
      "9   109   0.000364   0.235242  7.643935e-01\n",
      "10  110   0.000076   0.075360  9.245642e-01\n",
      "11  111   0.997202   0.002798  1.343153e-07\n",
      "12  112   0.052874   0.876635  7.049178e-02\n",
      "13  113   0.000059   0.068381  9.315599e-01\n",
      "14  114   0.986298   0.013698  3.236879e-06\n",
      "15  115   0.002562   0.562420  4.350178e-01\n",
      "16  116   0.054271   0.860268  8.546133e-02\n",
      "17  117   0.981537   0.018459  3.920587e-06\n",
      "18  118   0.989732   0.010266  1.793094e-06\n",
      "19  119   0.990169   0.009830  1.572438e-06\n",
      "20  120   0.015455   0.880904  1.036411e-01\n",
      "21  121   0.021652   0.883587  9.476057e-02\n",
      "22  122   0.001787   0.379847  6.183655e-01\n",
      "23  123   0.000065   0.048006  9.519287e-01\n",
      "24  124   0.000064   0.086036  9.138998e-01\n",
      "25  125   0.988903   0.011096  1.760461e-06\n",
      "26  126   0.000263   0.172666  8.270702e-01\n",
      "27  127   0.000510   0.216513  7.829766e-01\n",
      "28  128   0.000242   0.165022  8.347359e-01\n",
      "29  129   0.988183   0.011815  2.264432e-06\n",
      "30  130   0.024128   0.876179  9.969273e-02\n",
      "31  131   0.984830   0.015165  4.812646e-06\n",
      "32  132   0.001761   0.407811  5.904279e-01\n",
      "33  133   0.016058   0.807370  1.765722e-01\n",
      "34  134   0.007248   0.608790  3.839626e-01\n",
      "35  135   0.975645   0.024346  9.249625e-06\n",
      "36  136   0.000052   0.053163  9.467849e-01\n",
      "37  137   0.005661   0.692550  3.017898e-01\n",
      "38  138   0.985135   0.014862  3.423997e-06\n",
      "39  139   0.000923   0.157891  8.411858e-01\n",
      "40  140   0.977880   0.022111  8.839746e-06\n",
      "41  141   0.000120   0.112656  8.872238e-01\n",
      "42  142   0.000019   0.047257  9.527240e-01\n",
      "43  143   0.000245   0.125629  8.741255e-01\n",
      "44  144   0.974006   0.025987  7.587295e-06\n",
      "45  145   0.022328   0.782746  1.949266e-01\n",
      "46  146   0.109582   0.846037  4.438064e-02\n",
      "47  147   0.014828   0.717825  2.673467e-01\n",
      "48  148   0.013834   0.772525  2.136402e-01\n",
      "49  149   0.990233   0.009765  1.911912e-06\n",
      "50  150   0.960451   0.039526  2.336022e-05\n"
     ]
    }
   ],
   "source": [
    "# Generate Kaggle submit file\n",
    "\n",
    "# Encode feature vector\n",
    "df_test = pd.read_csv(\"https://data.heatonresearch.com/data/t81-558/datasets/kaggle_iris_test.csv\",\n",
    "                      na_values=['NA','?'])\n",
    "\n",
    "# Convert to numpy - Classification\n",
    "ids = df_test['id']\n",
    "df_test.drop('id', axis=1, inplace=True)\n",
    "x = df_test[['sepal_l', 'sepal_w', 'petal_l', 'petal_w']].values\n",
    "y = dummies.values\n",
    "\n",
    "# Generate predictions\n",
    "pred = model.predict(x)\n",
    "#pred\n",
    "\n",
    "# Create submission data set\n",
    "\n",
    "df_submit = pd.DataFrame(pred)\n",
    "df_submit.insert(0,'id',ids)\n",
    "df_submit.columns = ['id','species-0','species-1','species-2']\n",
    "\n",
    "df_submit.to_csv(\"iris_submit.csv\", index=False) # Write submit file locally\n",
    "\n",
    "print(df_submit)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### MPG as a Kaggle Competition (Regression)\n",
    "\n",
    "If the Auto MPG data were used as a Kaggle, you would be given the following three files:\n",
    "\n",
    "* [kaggle_mpg_test.csv](https://data.heatonresearch.com/data/t81-558/datasets/kaggle_auto_test.csv) - The data that Kaggle will evaluate you on.  Contains only input, you must provide answers.  (contains x)\n",
    "* [kaggle_mpg_train.csv](https://data.heatonresearch.com/data/t81-558/datasets/kaggle_auto_test.csv) - The data that you will use to train. (contains x and y)\n",
    "* [kaggle_mpg_sample.csv](https://data.heatonresearch.com/data/t81-558/datasets/kaggle_auto_sample.csv) - A sample submission for Kaggle. (contains x and y)\n",
    "\n",
    "Important features of the Kaggle iris files (that differ from how we've previously seen files):\n",
    "\n",
    "The following program generates a submission file for \"MPG Kaggle\".  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 261 samples, validate on 88 samples\n",
      "Epoch 1/1000\n",
      "261/261 - 0s - loss: 57911.6912 - val_loss: 24465.1060\n",
      "Epoch 2/1000\n",
      "261/261 - 0s - loss: 14306.0091 - val_loss: 3008.5809\n",
      "Epoch 3/1000\n",
      "261/261 - 0s - loss: 1132.2912 - val_loss: 468.8732\n",
      "Epoch 4/1000\n",
      "261/261 - 0s - loss: 1021.1585 - val_loss: 1452.0326\n",
      "Epoch 5/1000\n",
      "261/261 - 0s - loss: 1362.6016 - val_loss: 799.4204\n",
      "Epoch 6/1000\n",
      "261/261 - 0s - loss: 555.7693 - val_loss: 304.9727\n",
      "Epoch 7/1000\n",
      "261/261 - 0s - loss: 303.6679 - val_loss: 310.3975\n",
      "Epoch 8/1000\n",
      "261/261 - 0s - loss: 317.3761 - val_loss: 296.4635\n",
      "Epoch 9/1000\n",
      "261/261 - 0s - loss: 283.5167 - val_loss: 252.1378\n",
      "Epoch 10/1000\n",
      "261/261 - 0s - loss: 252.0144 - val_loss: 244.2337\n",
      "Epoch 11/1000\n",
      "261/261 - 0s - loss: 250.3846 - val_loss: 244.8350\n",
      "Epoch 12/1000\n",
      "261/261 - 0s - loss: 247.3224 - val_loss: 241.5592\n",
      "Epoch 13/1000\n",
      "261/261 - 0s - loss: 245.5661 - val_loss: 242.9108\n",
      "Epoch 14/1000\n",
      "261/261 - 0s - loss: 245.4319 - val_loss: 241.9302\n",
      "Epoch 15/1000\n",
      "261/261 - 0s - loss: 243.2965 - val_loss: 238.0927\n",
      "Epoch 16/1000\n",
      "261/261 - 0s - loss: 241.0569 - val_loss: 237.4987\n",
      "Epoch 17/1000\n",
      "261/261 - 0s - loss: 242.2063 - val_loss: 236.4684\n",
      "Epoch 18/1000\n",
      "261/261 - 0s - loss: 238.9881 - val_loss: 234.9435\n",
      "Epoch 19/1000\n",
      "261/261 - 0s - loss: 237.6926 - val_loss: 233.9827\n",
      "Epoch 20/1000\n",
      "261/261 - 0s - loss: 236.2662 - val_loss: 232.2227\n",
      "Epoch 21/1000\n",
      "261/261 - 0s - loss: 234.8052 - val_loss: 230.4309\n",
      "Epoch 22/1000\n",
      "261/261 - 0s - loss: 233.2169 - val_loss: 230.5371\n",
      "Epoch 23/1000\n",
      "261/261 - 0s - loss: 232.8251 - val_loss: 228.4490\n",
      "Epoch 24/1000\n",
      "261/261 - 0s - loss: 231.9342 - val_loss: 226.1372\n",
      "Epoch 25/1000\n",
      "261/261 - 0s - loss: 229.0155 - val_loss: 224.9210\n",
      "Epoch 26/1000\n",
      "261/261 - 0s - loss: 230.1828 - val_loss: 226.9475\n",
      "Epoch 27/1000\n",
      "261/261 - 0s - loss: 228.4033 - val_loss: 222.4413\n",
      "Epoch 28/1000\n",
      "261/261 - 0s - loss: 226.2988 - val_loss: 220.7109\n",
      "Epoch 29/1000\n",
      "261/261 - 0s - loss: 224.5981 - val_loss: 218.6426\n",
      "Epoch 30/1000\n",
      "261/261 - 0s - loss: 223.3171 - val_loss: 217.0329\n",
      "Epoch 31/1000\n",
      "261/261 - 0s - loss: 223.9491 - val_loss: 217.1737\n",
      "Epoch 32/1000\n",
      "261/261 - 0s - loss: 219.8517 - val_loss: 213.9108\n",
      "Epoch 33/1000\n",
      "261/261 - 0s - loss: 217.8795 - val_loss: 212.1759\n",
      "Epoch 34/1000\n",
      "261/261 - 0s - loss: 215.6078 - val_loss: 211.1876\n",
      "Epoch 35/1000\n",
      "261/261 - 0s - loss: 213.8885 - val_loss: 209.0320\n",
      "Epoch 36/1000\n",
      "261/261 - 0s - loss: 212.9289 - val_loss: 207.3585\n",
      "Epoch 37/1000\n",
      "261/261 - 0s - loss: 211.0405 - val_loss: 205.7438\n",
      "Epoch 38/1000\n",
      "261/261 - 0s - loss: 209.1992 - val_loss: 204.5727\n",
      "Epoch 39/1000\n",
      "261/261 - 0s - loss: 208.1457 - val_loss: 202.4382\n",
      "Epoch 40/1000\n",
      "261/261 - 0s - loss: 207.4388 - val_loss: 201.6114\n",
      "Epoch 41/1000\n",
      "261/261 - 0s - loss: 206.9847 - val_loss: 199.1185\n",
      "Epoch 42/1000\n",
      "261/261 - 0s - loss: 203.7296 - val_loss: 198.2328\n",
      "Epoch 43/1000\n",
      "261/261 - 0s - loss: 201.7638 - val_loss: 195.6919\n",
      "Epoch 44/1000\n",
      "261/261 - 0s - loss: 200.6981 - val_loss: 196.1719\n",
      "Epoch 45/1000\n",
      "261/261 - 0s - loss: 199.0079 - val_loss: 192.5120\n",
      "Epoch 46/1000\n",
      "261/261 - 0s - loss: 196.7346 - val_loss: 190.9500\n",
      "Epoch 47/1000\n",
      "261/261 - 0s - loss: 194.1735 - val_loss: 188.8801\n",
      "Epoch 48/1000\n",
      "261/261 - 0s - loss: 193.1979 - val_loss: 187.3858\n",
      "Epoch 49/1000\n",
      "261/261 - 0s - loss: 200.1771 - val_loss: 191.0603\n",
      "Epoch 50/1000\n",
      "261/261 - 0s - loss: 189.7782 - val_loss: 184.3938\n",
      "Epoch 51/1000\n",
      "261/261 - 0s - loss: 188.4902 - val_loss: 182.1200\n",
      "Epoch 52/1000\n",
      "261/261 - 0s - loss: 186.4416 - val_loss: 180.2011\n",
      "Epoch 53/1000\n",
      "261/261 - 0s - loss: 184.8636 - val_loss: 178.4774\n",
      "Epoch 54/1000\n",
      "261/261 - 0s - loss: 183.2283 - val_loss: 178.7142\n",
      "Epoch 55/1000\n",
      "261/261 - 0s - loss: 180.8820 - val_loss: 175.2477\n",
      "Epoch 56/1000\n",
      "261/261 - 0s - loss: 181.9741 - val_loss: 173.9610\n",
      "Epoch 57/1000\n",
      "261/261 - 0s - loss: 178.0085 - val_loss: 172.1204\n",
      "Epoch 58/1000\n",
      "261/261 - 0s - loss: 177.6698 - val_loss: 172.5644\n",
      "Epoch 59/1000\n",
      "261/261 - 0s - loss: 176.2773 - val_loss: 168.2883\n",
      "Epoch 60/1000\n",
      "261/261 - 0s - loss: 173.1984 - val_loss: 167.7326\n",
      "Epoch 61/1000\n",
      "261/261 - 0s - loss: 171.9081 - val_loss: 164.9589\n",
      "Epoch 62/1000\n",
      "261/261 - 0s - loss: 171.2389 - val_loss: 163.9004\n",
      "Epoch 63/1000\n",
      "261/261 - 0s - loss: 169.1427 - val_loss: 162.1389\n",
      "Epoch 64/1000\n",
      "261/261 - 0s - loss: 167.0260 - val_loss: 160.9548\n",
      "Epoch 65/1000\n",
      "261/261 - 0s - loss: 165.2057 - val_loss: 158.2418\n",
      "Epoch 66/1000\n",
      "261/261 - 0s - loss: 163.7387 - val_loss: 157.6366\n",
      "Epoch 67/1000\n",
      "261/261 - 0s - loss: 162.4386 - val_loss: 155.1394\n",
      "Epoch 68/1000\n",
      "261/261 - 0s - loss: 160.4362 - val_loss: 153.4851\n",
      "Epoch 69/1000\n",
      "261/261 - 0s - loss: 159.7696 - val_loss: 151.6310\n",
      "Epoch 70/1000\n",
      "261/261 - 0s - loss: 156.8633 - val_loss: 154.0878\n",
      "Epoch 71/1000\n",
      "261/261 - 0s - loss: 159.7552 - val_loss: 149.2042\n",
      "Epoch 72/1000\n",
      "261/261 - 0s - loss: 154.3063 - val_loss: 147.3970\n",
      "Epoch 73/1000\n",
      "261/261 - 0s - loss: 153.2741 - val_loss: 145.1558\n",
      "Epoch 74/1000\n",
      "261/261 - 0s - loss: 154.2913 - val_loss: 145.6223\n",
      "Epoch 75/1000\n",
      "261/261 - 0s - loss: 152.3954 - val_loss: 144.7565\n",
      "Epoch 76/1000\n",
      "261/261 - 0s - loss: 150.0854 - val_loss: 142.9019\n",
      "Epoch 77/1000\n",
      "261/261 - 0s - loss: 147.8938 - val_loss: 138.5884\n",
      "Epoch 78/1000\n",
      "261/261 - 0s - loss: 144.5425 - val_loss: 137.3837\n",
      "Epoch 79/1000\n",
      "261/261 - 0s - loss: 144.0798 - val_loss: 135.4681\n",
      "Epoch 80/1000\n",
      "261/261 - 0s - loss: 141.7498 - val_loss: 133.9645\n",
      "Epoch 81/1000\n",
      "261/261 - 0s - loss: 139.7431 - val_loss: 132.5269\n",
      "Epoch 82/1000\n",
      "261/261 - 0s - loss: 139.5900 - val_loss: 131.4013\n",
      "Epoch 83/1000\n",
      "261/261 - 0s - loss: 137.0045 - val_loss: 129.3308\n",
      "Epoch 84/1000\n",
      "261/261 - 0s - loss: 135.1712 - val_loss: 127.9085\n",
      "Epoch 85/1000\n",
      "261/261 - 0s - loss: 133.9908 - val_loss: 126.6167\n",
      "Epoch 86/1000\n",
      "261/261 - 0s - loss: 132.7881 - val_loss: 125.0145\n",
      "Epoch 87/1000\n",
      "261/261 - 0s - loss: 132.3677 - val_loss: 124.4792\n",
      "Epoch 88/1000\n",
      "261/261 - 0s - loss: 129.0288 - val_loss: 121.8587\n",
      "Epoch 89/1000\n",
      "261/261 - 0s - loss: 127.9998 - val_loss: 120.4485\n",
      "Epoch 90/1000\n",
      "261/261 - 0s - loss: 126.5930 - val_loss: 119.1081\n",
      "Epoch 91/1000\n",
      "261/261 - 0s - loss: 124.9949 - val_loss: 117.4116\n",
      "Epoch 92/1000\n",
      "261/261 - 0s - loss: 123.7998 - val_loss: 116.1554\n",
      "Epoch 93/1000\n",
      "261/261 - 0s - loss: 125.4202 - val_loss: 114.6380\n",
      "Epoch 94/1000\n",
      "261/261 - 0s - loss: 126.5583 - val_loss: 115.3558\n",
      "Epoch 95/1000\n",
      "261/261 - 0s - loss: 119.3499 - val_loss: 113.3887\n",
      "Epoch 96/1000\n",
      "261/261 - 0s - loss: 124.9014 - val_loss: 112.6252\n",
      "Epoch 97/1000\n",
      "261/261 - 0s - loss: 115.9393 - val_loss: 111.2288\n",
      "Epoch 98/1000\n",
      "261/261 - 0s - loss: 117.8388 - val_loss: 108.5923\n",
      "Epoch 99/1000\n",
      "261/261 - 0s - loss: 113.2265 - val_loss: 106.4497\n",
      "Epoch 100/1000\n",
      "261/261 - 0s - loss: 121.7596 - val_loss: 104.4762\n",
      "Epoch 101/1000\n",
      "261/261 - 0s - loss: 112.7481 - val_loss: 105.4822\n",
      "Epoch 102/1000\n",
      "261/261 - 0s - loss: 113.0200 - val_loss: 101.7701\n",
      "Epoch 103/1000\n",
      "261/261 - 0s - loss: 115.4485 - val_loss: 103.4175\n",
      "Epoch 104/1000\n",
      "261/261 - 0s - loss: 109.0804 - val_loss: 101.5245\n",
      "Epoch 105/1000\n",
      "261/261 - 0s - loss: 107.6733 - val_loss: 99.0502\n",
      "Epoch 106/1000\n",
      "261/261 - 0s - loss: 107.2623 - val_loss: 97.1591\n",
      "Epoch 107/1000\n",
      "261/261 - 0s - loss: 103.5797 - val_loss: 95.6571\n",
      "Epoch 108/1000\n",
      "261/261 - 0s - loss: 102.2904 - val_loss: 94.9382\n",
      "Epoch 109/1000\n",
      "261/261 - 0s - loss: 102.2591 - val_loss: 93.1886\n",
      "Epoch 110/1000\n",
      "261/261 - 0s - loss: 100.0956 - val_loss: 92.0002\n",
      "Epoch 111/1000\n",
      "261/261 - 0s - loss: 98.9123 - val_loss: 90.7469\n",
      "Epoch 112/1000\n",
      "261/261 - 0s - loss: 97.6304 - val_loss: 89.7667\n",
      "Epoch 113/1000\n",
      "261/261 - 0s - loss: 96.4284 - val_loss: 89.0412\n",
      "Epoch 114/1000\n",
      "261/261 - 0s - loss: 95.6745 - val_loss: 87.4017\n",
      "Epoch 115/1000\n",
      "261/261 - 0s - loss: 94.0422 - val_loss: 86.0830\n",
      "Epoch 116/1000\n",
      "261/261 - 0s - loss: 93.1400 - val_loss: 84.9548\n",
      "Epoch 117/1000\n",
      "261/261 - 0s - loss: 91.6857 - val_loss: 84.0395\n",
      "Epoch 118/1000\n",
      "261/261 - 0s - loss: 90.8716 - val_loss: 83.0955\n",
      "Epoch 119/1000\n",
      "261/261 - 0s - loss: 89.4472 - val_loss: 81.5549\n",
      "Epoch 120/1000\n",
      "261/261 - 0s - loss: 90.3298 - val_loss: 80.8134\n",
      "Epoch 121/1000\n",
      "261/261 - 0s - loss: 88.5368 - val_loss: 80.5562\n",
      "Epoch 122/1000\n",
      "261/261 - 0s - loss: 88.2283 - val_loss: 79.1708\n",
      "Epoch 123/1000\n",
      "261/261 - 0s - loss: 85.5158 - val_loss: 77.2081\n",
      "Epoch 124/1000\n",
      "261/261 - 0s - loss: 84.6625 - val_loss: 77.1045\n",
      "Epoch 125/1000\n",
      "261/261 - 0s - loss: 83.9567 - val_loss: 75.2061\n",
      "Epoch 126/1000\n",
      "261/261 - 0s - loss: 82.5359 - val_loss: 74.4080\n",
      "Epoch 127/1000\n",
      "261/261 - 0s - loss: 84.3972 - val_loss: 73.8694\n",
      "Epoch 128/1000\n",
      "261/261 - 0s - loss: 83.8952 - val_loss: 72.2010\n",
      "Epoch 129/1000\n",
      "261/261 - 0s - loss: 82.8451 - val_loss: 74.4700\n",
      "Epoch 130/1000\n",
      "261/261 - 0s - loss: 79.3057 - val_loss: 71.2101\n",
      "Epoch 131/1000\n",
      "261/261 - 0s - loss: 78.0721 - val_loss: 69.3212\n",
      "Epoch 132/1000\n",
      "261/261 - 0s - loss: 76.8455 - val_loss: 68.3555\n",
      "Epoch 133/1000\n",
      "261/261 - 0s - loss: 78.8544 - val_loss: 67.3812\n",
      "Epoch 134/1000\n",
      "261/261 - 0s - loss: 75.6548 - val_loss: 66.3610\n",
      "Epoch 135/1000\n",
      "261/261 - 0s - loss: 74.6969 - val_loss: 66.6813\n",
      "Epoch 136/1000\n",
      "261/261 - 0s - loss: 73.8797 - val_loss: 64.8747\n",
      "Epoch 137/1000\n",
      "261/261 - 0s - loss: 74.1671 - val_loss: 64.7925\n",
      "Epoch 138/1000\n",
      "261/261 - 0s - loss: 73.1038 - val_loss: 63.3704\n",
      "Epoch 139/1000\n",
      "261/261 - 0s - loss: 71.4968 - val_loss: 63.8811\n",
      "Epoch 140/1000\n",
      "261/261 - 0s - loss: 69.9367 - val_loss: 61.5494\n",
      "Epoch 141/1000\n",
      "261/261 - 0s - loss: 70.0337 - val_loss: 60.5077\n",
      "Epoch 142/1000\n",
      "261/261 - 0s - loss: 67.8479 - val_loss: 59.4890\n",
      "Epoch 143/1000\n",
      "261/261 - 0s - loss: 66.7949 - val_loss: 58.9073\n",
      "Epoch 144/1000\n",
      "261/261 - 0s - loss: 66.1649 - val_loss: 57.9403\n",
      "Epoch 145/1000\n",
      "261/261 - 0s - loss: 66.0125 - val_loss: 58.3263\n",
      "Epoch 146/1000\n",
      "261/261 - 0s - loss: 65.4468 - val_loss: 56.4226\n",
      "Epoch 147/1000\n",
      "261/261 - 0s - loss: 65.4101 - val_loss: 56.0215\n",
      "Epoch 148/1000\n",
      "261/261 - 0s - loss: 62.8652 - val_loss: 55.1651\n",
      "Epoch 149/1000\n",
      "261/261 - 0s - loss: 61.9235 - val_loss: 55.1786\n",
      "Epoch 150/1000\n",
      "261/261 - 0s - loss: 61.3767 - val_loss: 54.1903\n",
      "Epoch 151/1000\n",
      "261/261 - 0s - loss: 61.4330 - val_loss: 53.0662\n",
      "Epoch 152/1000\n",
      "261/261 - 0s - loss: 60.7498 - val_loss: 52.2158\n",
      "Epoch 153/1000\n",
      "261/261 - 0s - loss: 59.4143 - val_loss: 51.5297\n",
      "Epoch 154/1000\n",
      "261/261 - 0s - loss: 59.9519 - val_loss: 50.9271\n",
      "Epoch 155/1000\n",
      "261/261 - 0s - loss: 59.3391 - val_loss: 50.6017\n",
      "Epoch 156/1000\n",
      "261/261 - 0s - loss: 57.8237 - val_loss: 49.5864\n",
      "Epoch 157/1000\n",
      "261/261 - 0s - loss: 56.9233 - val_loss: 49.0480\n",
      "Epoch 158/1000\n",
      "261/261 - 0s - loss: 55.9277 - val_loss: 48.5914\n",
      "Epoch 159/1000\n",
      "261/261 - 0s - loss: 55.9890 - val_loss: 47.9270\n",
      "Epoch 160/1000\n",
      "261/261 - 0s - loss: 54.8821 - val_loss: 47.2261\n",
      "Epoch 161/1000\n",
      "261/261 - 0s - loss: 54.3646 - val_loss: 46.9130\n",
      "Epoch 162/1000\n",
      "261/261 - 0s - loss: 55.1221 - val_loss: 46.3666\n",
      "Epoch 163/1000\n",
      "261/261 - 0s - loss: 56.4017 - val_loss: 45.3519\n",
      "Epoch 164/1000\n",
      "261/261 - 0s - loss: 53.9293 - val_loss: 44.9182\n",
      "Epoch 165/1000\n",
      "261/261 - 0s - loss: 52.8715 - val_loss: 44.2564\n",
      "Epoch 166/1000\n",
      "261/261 - 0s - loss: 51.2135 - val_loss: 43.5814\n",
      "Epoch 167/1000\n",
      "261/261 - 0s - loss: 50.4227 - val_loss: 43.4569\n",
      "Epoch 168/1000\n",
      "261/261 - 0s - loss: 50.4024 - val_loss: 42.5375\n",
      "Epoch 169/1000\n",
      "261/261 - 0s - loss: 49.6260 - val_loss: 42.0628\n",
      "Epoch 170/1000\n",
      "261/261 - 0s - loss: 49.0483 - val_loss: 41.5198\n",
      "Epoch 171/1000\n",
      "261/261 - 0s - loss: 48.6025 - val_loss: 41.0386\n",
      "Epoch 172/1000\n",
      "261/261 - 0s - loss: 48.0345 - val_loss: 40.5795\n",
      "Epoch 173/1000\n",
      "261/261 - 0s - loss: 47.7011 - val_loss: 40.1023\n",
      "Epoch 174/1000\n",
      "261/261 - 0s - loss: 47.2931 - val_loss: 39.6843\n",
      "Epoch 175/1000\n",
      "261/261 - 0s - loss: 46.7045 - val_loss: 39.2973\n",
      "Epoch 176/1000\n",
      "261/261 - 0s - loss: 47.6070 - val_loss: 39.1371\n",
      "Epoch 177/1000\n",
      "261/261 - 0s - loss: 46.0122 - val_loss: 38.3189\n",
      "Epoch 178/1000\n",
      "261/261 - 0s - loss: 45.7323 - val_loss: 38.0894\n",
      "Epoch 179/1000\n",
      "261/261 - 0s - loss: 44.9723 - val_loss: 37.8146\n",
      "Epoch 180/1000\n",
      "261/261 - 0s - loss: 48.0190 - val_loss: 39.2085\n",
      "Epoch 181/1000\n",
      "261/261 - 0s - loss: 48.6329 - val_loss: 37.4405\n",
      "Epoch 182/1000\n",
      "261/261 - 0s - loss: 48.3010 - val_loss: 36.9400\n",
      "Epoch 183/1000\n",
      "261/261 - 0s - loss: 43.4618 - val_loss: 36.4700\n",
      "Epoch 184/1000\n",
      "261/261 - 0s - loss: 44.0408 - val_loss: 35.6435\n",
      "Epoch 185/1000\n",
      "261/261 - 0s - loss: 42.3352 - val_loss: 35.0616\n",
      "Epoch 186/1000\n",
      "261/261 - 0s - loss: 44.4314 - val_loss: 36.2616\n",
      "Epoch 187/1000\n",
      "261/261 - 0s - loss: 49.2686 - val_loss: 36.4352\n",
      "Epoch 188/1000\n",
      "261/261 - 0s - loss: 46.2653 - val_loss: 35.5499\n",
      "Epoch 189/1000\n",
      "261/261 - 0s - loss: 43.2500 - val_loss: 35.1195\n",
      "Epoch 190/1000\n",
      "261/261 - 0s - loss: 43.6424 - val_loss: 34.7520\n",
      "Epoch 191/1000\n",
      "261/261 - 0s - loss: 41.3217 - val_loss: 34.1096\n",
      "Epoch 192/1000\n",
      "261/261 - 0s - loss: 40.3076 - val_loss: 33.2629\n",
      "Epoch 193/1000\n",
      "261/261 - 0s - loss: 39.5597 - val_loss: 32.6054\n",
      "Epoch 194/1000\n",
      "261/261 - 0s - loss: 39.4696 - val_loss: 33.4332\n",
      "Epoch 195/1000\n",
      "261/261 - 0s - loss: 39.9985 - val_loss: 32.7632\n",
      "Epoch 196/1000\n",
      "261/261 - 0s - loss: 40.6351 - val_loss: 32.1845\n",
      "Epoch 197/1000\n",
      "261/261 - 0s - loss: 38.5683 - val_loss: 31.7487\n",
      "Epoch 198/1000\n",
      "261/261 - 0s - loss: 39.1509 - val_loss: 31.3765\n",
      "Epoch 199/1000\n",
      "261/261 - 0s - loss: 37.6904 - val_loss: 31.0275\n",
      "Epoch 200/1000\n",
      "261/261 - 0s - loss: 37.5366 - val_loss: 31.0889\n",
      "Epoch 201/1000\n",
      "261/261 - 0s - loss: 37.8550 - val_loss: 30.6194\n",
      "Epoch 202/1000\n",
      "261/261 - 0s - loss: 38.0003 - val_loss: 30.7284\n",
      "Epoch 203/1000\n",
      "261/261 - 0s - loss: 38.2464 - val_loss: 31.1378\n",
      "Epoch 204/1000\n",
      "261/261 - 0s - loss: 37.7872 - val_loss: 29.8230\n",
      "Epoch 205/1000\n",
      "261/261 - 0s - loss: 36.2347 - val_loss: 29.6210\n",
      "Epoch 206/1000\n",
      "261/261 - 0s - loss: 36.2658 - val_loss: 30.1315\n",
      "Epoch 207/1000\n",
      "261/261 - 0s - loss: 37.2058 - val_loss: 29.5572\n",
      "Epoch 208/1000\n",
      "261/261 - 0s - loss: 36.0193 - val_loss: 29.3442\n",
      "Epoch 209/1000\n",
      "261/261 - 0s - loss: 35.4656 - val_loss: 28.9099\n",
      "Epoch 210/1000\n",
      "261/261 - 0s - loss: 35.2463 - val_loss: 29.2081\n",
      "Epoch 211/1000\n",
      "261/261 - 0s - loss: 34.9982 - val_loss: 28.5349\n",
      "Epoch 212/1000\n",
      "261/261 - 0s - loss: 34.7378 - val_loss: 28.3040\n",
      "Epoch 213/1000\n",
      "261/261 - 0s - loss: 34.9454 - val_loss: 28.1356\n",
      "Epoch 214/1000\n",
      "261/261 - 0s - loss: 34.6790 - val_loss: 29.0780\n",
      "Epoch 215/1000\n",
      "261/261 - 0s - loss: 37.1930 - val_loss: 28.5673\n",
      "Epoch 216/1000\n",
      "261/261 - 0s - loss: 33.9258 - val_loss: 27.7105\n",
      "Epoch 217/1000\n",
      "261/261 - 0s - loss: 33.6924 - val_loss: 27.3721\n",
      "Epoch 218/1000\n",
      "261/261 - 0s - loss: 34.3994 - val_loss: 27.7282\n",
      "Epoch 219/1000\n",
      "261/261 - 0s - loss: 33.8463 - val_loss: 27.5516\n",
      "Epoch 220/1000\n",
      "261/261 - 0s - loss: 33.4583 - val_loss: 27.1268\n",
      "Epoch 221/1000\n",
      "261/261 - 0s - loss: 33.0171 - val_loss: 26.9615\n",
      "Epoch 222/1000\n",
      "261/261 - 0s - loss: 32.9557 - val_loss: 26.8393\n",
      "Epoch 223/1000\n",
      "261/261 - 0s - loss: 33.5171 - val_loss: 27.7535\n",
      "Epoch 224/1000\n",
      "261/261 - 0s - loss: 34.1597 - val_loss: 29.1331\n",
      "Epoch 225/1000\n",
      "261/261 - 0s - loss: 35.9033 - val_loss: 28.4633\n",
      "Epoch 226/1000\n",
      "261/261 - 0s - loss: 33.5818 - val_loss: 26.1210\n",
      "Epoch 227/1000\n",
      "261/261 - 0s - loss: 32.2207 - val_loss: 25.9743\n",
      "Epoch 228/1000\n",
      "261/261 - 0s - loss: 31.6236 - val_loss: 26.1432\n",
      "Epoch 229/1000\n",
      "261/261 - 0s - loss: 32.2752 - val_loss: 25.8567\n",
      "Epoch 230/1000\n",
      "261/261 - 0s - loss: 32.2266 - val_loss: 26.5454\n",
      "Epoch 231/1000\n",
      "261/261 - 0s - loss: 33.6443 - val_loss: 28.4963\n",
      "Epoch 232/1000\n",
      "261/261 - 0s - loss: 34.2461 - val_loss: 28.3813\n",
      "Epoch 233/1000\n",
      "261/261 - 0s - loss: 32.4732 - val_loss: 26.3547\n",
      "Epoch 234/1000\n",
      "261/261 - 0s - loss: 31.6554 - val_loss: 25.1999\n",
      "Epoch 235/1000\n",
      "261/261 - 0s - loss: 32.4069 - val_loss: 26.6240\n",
      "Epoch 236/1000\n",
      "261/261 - 0s - loss: 32.0460 - val_loss: 25.3677\n",
      "Epoch 237/1000\n",
      "261/261 - 0s - loss: 31.2378 - val_loss: 25.5845\n",
      "Epoch 238/1000\n",
      "261/261 - 0s - loss: 30.7541 - val_loss: 24.8422\n",
      "Epoch 239/1000\n",
      "261/261 - 0s - loss: 30.4288 - val_loss: 24.8658\n",
      "Epoch 240/1000\n",
      "261/261 - 0s - loss: 30.4009 - val_loss: 24.8365\n",
      "Epoch 241/1000\n",
      "261/261 - 0s - loss: 30.2608 - val_loss: 25.2763\n",
      "Epoch 242/1000\n",
      "261/261 - 0s - loss: 32.3225 - val_loss: 27.6487\n",
      "Epoch 243/1000\n",
      "261/261 - 0s - loss: 30.9560 - val_loss: 25.1304\n",
      "Epoch 244/1000\n",
      "261/261 - 0s - loss: 30.7176 - val_loss: 24.3037\n",
      "Epoch 245/1000\n",
      "261/261 - 0s - loss: 30.0964 - val_loss: 24.1632\n",
      "Epoch 246/1000\n",
      "261/261 - 0s - loss: 29.7507 - val_loss: 24.1421\n",
      "Epoch 247/1000\n",
      "261/261 - 0s - loss: 30.8254 - val_loss: 24.3073\n",
      "Epoch 248/1000\n",
      "261/261 - 0s - loss: 29.8273 - val_loss: 25.6336\n",
      "Epoch 249/1000\n",
      "261/261 - 0s - loss: 31.0508 - val_loss: 24.3445\n",
      "Epoch 250/1000\n",
      "261/261 - 0s - loss: 34.7534 - val_loss: 32.5582\n",
      "Epoch 251/1000\n",
      "Restoring model weights from the end of the best epoch.\n",
      "261/261 - 0s - loss: 36.6671 - val_loss: 30.6870\n",
      "Epoch 00251: early stopping\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense, Activation\n",
    "from sklearn.model_selection import train_test_split\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "import pandas as pd\n",
    "import io\n",
    "import os\n",
    "import requests\n",
    "import numpy as np\n",
    "from sklearn import metrics\n",
    "\n",
    "save_path = \".\"\n",
    "\n",
    "df = pd.read_csv(\n",
    "    \"https://data.heatonresearch.com/data/t81-558/datasets/kaggle_auto_train.csv\", \n",
    "    na_values=['NA', '?'])\n",
    "\n",
    "cars = df['name']\n",
    "\n",
    "# Handle missing value\n",
    "df['horsepower'] = df['horsepower'].fillna(df['horsepower'].median())\n",
    "\n",
    "# Pandas to Numpy\n",
    "x = df[['cylinders', 'displacement', 'horsepower', 'weight',\n",
    "       'acceleration', 'year', 'origin']].values\n",
    "y = df['mpg'].values # regression\n",
    "\n",
    "# Split into train/test\n",
    "x_train, x_test, y_train, y_test = train_test_split(    \n",
    "    x, y, test_size=0.25, random_state=42)\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, patience=5, \n",
    "                        verbose=1, mode='auto', restore_best_weights=True)\n",
    "model.fit(x_train,y_train,validation_data=(x_test,y_test),\n",
    "          verbose=2,callbacks=[monitor],epochs=1000)\n",
    "\n",
    "# Predict\n",
    "pred = model.predict(x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Final score (RMSE): 4.913462818455655\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": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "     id        mpg\n",
      "0   350  27.541994\n",
      "1   351  25.862688\n",
      "2   352  26.286074\n",
      "3   353  28.067247\n",
      "4   354  26.922823\n",
      "5   355  29.997942\n",
      "6   356  27.195421\n",
      "7   357  27.913530\n",
      "8   358  29.707598\n",
      "9   359  27.877825\n",
      "10  360  28.481096\n",
      "11  361  27.393724\n",
      "12  362  27.112352\n",
      "13  363  29.838976\n",
      "14  364  22.223558\n",
      "15  365  11.889268\n",
      "16  366  22.200274\n",
      "17  367  20.895983\n",
      "18  368  29.619097\n",
      "19  369  29.607958\n",
      "20  370  28.666338\n",
      "21  371  28.795122\n",
      "22  372  26.298723\n",
      "23  373  26.294970\n",
      "24  374  27.767382\n",
      "25  375  28.209459\n",
      "26  376  26.572680\n",
      "27  377  28.218096\n",
      "28  378  27.944262\n",
      "29  379  25.759920\n",
      "30  380  27.392458\n",
      "31  381  27.163240\n",
      "32  382  27.431566\n",
      "33  383  27.360979\n",
      "34  384  27.490450\n",
      "35  385  27.584368\n",
      "36  386  27.756899\n",
      "37  387  25.689104\n",
      "38  388  16.059099\n",
      "39  389  24.993456\n",
      "40  390  20.356920\n",
      "41  391  27.254869\n",
      "42  392  25.351488\n",
      "43  393  26.956682\n",
      "44  394  26.871660\n",
      "45  395  27.191133\n",
      "46  396  24.900072\n",
      "47  397  28.000689\n",
      "48  398  28.800188\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "# Generate Kaggle submit file\n",
    "\n",
    "# Encode feature vector\n",
    "df_test = pd.read_csv(\"https://data.heatonresearch.com/data/t81-558/datasets/kaggle_auto_test.csv\",\n",
    "                      na_values=['NA','?'])\n",
    "\n",
    "# Convert to numpy - regression\n",
    "ids = df_test['id']\n",
    "df_test.drop('id', axis=1, inplace=True)\n",
    "\n",
    "# Handle missing value\n",
    "df_test['horsepower'] = df_test['horsepower'].fillna(df['horsepower'].median())\n",
    "\n",
    "x = df_test[['cylinders', 'displacement', 'horsepower', 'weight',\n",
    "       'acceleration', 'year', 'origin']].values\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "# Generate predictions\n",
    "pred = model.predict(x)\n",
    "#pred\n",
    "\n",
    "# Create submission data set\n",
    "\n",
    "df_submit = pd.DataFrame(pred)\n",
    "df_submit.insert(0,'id',ids)\n",
    "df_submit.columns = ['id','mpg']\n",
    "\n",
    "df_submit.to_csv(\"auto_submit.csv\", index=False) # Write submit file locally\n",
    "\n",
    "print(df_submit)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Module 8 Assignment\n",
    "\n",
    "You can find the first assignment here: [assignment 8](https://github.com/jeffheaton/t81_558_deep_learning/blob/master/assignments/assignment_yourname_class8.ipynb)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python 3.6 (tensorflow-2.0)",
   "language": "python",
   "name": "tensorflow-2.0"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}