anom_detection_master.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.layers import Dense, Input, Conv2D, Conv2DTranspose, Lambda, Flatten, Reshape, MaxPooling2D\n",
    "from keras.models import Model\n",
    "from keras.losses import mean_squared_error, binary_crossentropy, kullback_leibler_divergence\n",
    "from keras.callbacks import EarlyStopping, TerminateOnNaN, ReduceLROnPlateau\n",
    "from keras import backend as K\n",
    "import tensorflow as tf\n",
    "from sklearn.preprocessing import StandardScaler#\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import os, scipy\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "from phenom_to_df import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Get the data\n",
    "df = csv_to_df('test/')\n",
    "df = df.fillna(0)\n",
    "variables = [entry for entry in df.columns if entry[0] == 'E'] + [entry for entry in df.columns if entry[0:2] == 'pt'] + [entry for entry in df.columns if entry[0:2] == 'et'] + [entry for entry in df.columns if entry[0:2] == 'ph']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Now we one_hot encode the data to make sure we have a class definition\n",
    "df = df[['process_ID']+variables]\n",
    "one_hot = pd.get_dummies(df['process_ID'])\n",
    "processes = one_hot.columns\n",
    "df.drop('process_ID', axis = 'columns', inplace = True)\n",
    "df = pd.concat([df, one_hot], sort = False, axis = 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "#We have created a df of our values and some kind of class label\n",
    "#We should normalize\n",
    "x = df[variables].values\n",
    "x_scaled = StandardScaler().fit_transform(x)\n",
    "df[variables] = x_scaled\n",
    "\n",
    "x_train, x_test, y_train, y_test = train_test_split(df[variables].values,\n",
    "\t\t\t\t\t\t    df[processes].values, \n",
    "                                                    shuffle = True,\n",
    "                                                    random_state = 42,\n",
    "                                                    test_size = 0.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Reshape the data. This will have to be done explicitly until you think of a clever way of doing it\n",
    "x_train = x_train.reshape((x_train.shape[0], 4, 3, 5))\n",
    "x_test = x_test.reshape((x_test.shape[0], 4, 3, 5))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def sampling(args):\n",
    "    z_mean, z_log_var=args\n",
    "    epsilon=tf.random.uniform(shape=(K.shape(z_mean)[0], latent_dim))\n",
    "    return z_mean+K.exp(z_log_var/2)*epsilon"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "#VAE Model\n",
    "original_dim = x_train.shape[0]\n",
    "input_shape = x_train.shape[1:]\n",
    "latent_dim = 4\n",
    "intermediate_dim = 50\n",
    "kernel_max_norm = 1000.\n",
    "act_fun = 'relu'\n",
    "epsilon_std = 1.\n",
    "filters = 32\n",
    "kernel_size = (3, 3)\n",
    "strides = (1, 1)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Layers\n",
    "x_input = Input(shape = input_shape, name = 'Input')\n",
    "conv_1 = Conv2D(filters = filters, \n",
    "                activation = act_fun,\n",
    "                kernel_size = kernel_size,\n",
    "                strides = strides, \n",
    "                padding = 'same')(x_input)\n",
    "pool_1 = MaxPooling2D(pool_size = (1, 1),\n",
    "                      strides = 2,\n",
    "                      data_format = 'channels_first')(conv_1)\n",
    "conv_2 = Conv2D(filters = filters, \n",
    "                activation = act_fun, \n",
    "                kernel_size = kernel_size, \n",
    "                strides = strides, \n",
    "                padding = 'same')(pool_1)\n",
    "\n",
    "#Shape info needed to inform the dense layer in the decoder\n",
    "flatten = Flatten()(conv_2)\n",
    "#cnn = Model(x_input, flatten, name = 'cnn')\n",
    "#cnn.summary()\n",
    "\n",
    "shape = input_shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"encoder\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "Input (InputLayer)              (None, 4, 3, 5)      0                                            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_1 (Conv2D)               (None, 4, 3, 32)     1472        Input[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2D)  (None, 4, 2, 16)     0           conv2d_1[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_2 (Conv2D)               (None, 4, 2, 32)     4640        max_pooling2d_1[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "flatten_1 (Flatten)             (None, 256)          0           conv2d_2[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_encoder_1 (Dense)         (None, 50)           12850       flatten_1[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_encoder_2 (Dense)         (None, 16)           816         dense_encoder_1[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "z_mean (Dense)                  (None, 4)            68          dense_encoder_2[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "z_var (Dense)                   (None, 4)            68          dense_encoder_2[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "sampling (Lambda)               (None, 4)            0           z_mean[0][0]                     \n",
      "                                                                 z_var[0][0]                      \n",
      "==================================================================================================\n",
      "Total params: 19,914\n",
      "Trainable params: 19,914\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "#Encoding\n",
    "dense_encoder_1 = Dense(intermediate_dim,\n",
    "                        activation = act_fun,\n",
    "                        name = 'dense_encoder_1')(flatten)\n",
    "dense_encoder_2 = Dense(16,\n",
    "                        activation = act_fun,\n",
    "                        name = 'dense_encoder_2')(dense_encoder_1)\n",
    "z_mean = Dense(latent_dim, name = 'z_mean')(dense_encoder_2)\n",
    "z_var = Dense(latent_dim, name = 'z_var')(dense_encoder_2)\n",
    "\n",
    "z = Lambda(sampling, output_shape = (latent_dim, ), name = 'sampling')([z_mean, z_var])\n",
    "\n",
    "encoder = Model(x_input, [z_mean, z_var, z], name = 'encoder')\n",
    "encoder.summary()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"decoder\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "decoder_input (InputLayer)   (None, 4)                 0         \n",
      "_________________________________________________________________\n",
      "dense_decoder_1 (Dense)      (None, 60)                300       \n",
      "_________________________________________________________________\n",
      "reshape_1 (Reshape)          (None, 4, 3, 5)           0         \n",
      "_________________________________________________________________\n",
      "deconv_1 (Conv2DTranspose)   (None, 4, 3, 32)          1472      \n",
      "_________________________________________________________________\n",
      "deconv_2 (Conv2DTranspose)   (None, 4, 3, 16)          4624      \n",
      "_________________________________________________________________\n",
      "decoder_output (Conv2DTransp (None, 4, 3, 5)           725       \n",
      "=================================================================\n",
      "Total params: 7,121\n",
      "Trainable params: 7,121\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "#Decoding\n",
    "decoder_input = Input(shape = (latent_dim, ), name = 'decoder_input')\n",
    "dense_decoder_1 = Dense(shape[0]*shape[1]*shape[2], \n",
    "                        activation = act_fun, name = 'dense_decoder_1')(decoder_input)\n",
    "reshape = Reshape((shape[0], shape[1], shape[2]))(dense_decoder_1)\n",
    "\n",
    "deconv_1 = Conv2DTranspose(filters = filters,\n",
    "                           kernel_size = kernel_size, \n",
    "                           activation = act_fun, \n",
    "                           strides = strides, \n",
    "                           padding = 'same', \n",
    "                           name = 'deconv_1')(reshape)\n",
    "deconv_2 = Conv2DTranspose(filters = 16,\n",
    "                           kernel_size = kernel_size, \n",
    "                           activation = act_fun, \n",
    "                           strides = strides, \n",
    "                           padding = 'same', \n",
    "                           name = 'deconv_2')(deconv_1)\n",
    "\n",
    "output = Conv2DTranspose(filters = 5, \n",
    "                         kernel_size = kernel_size, \n",
    "                         activation = 'sigmoid', \n",
    "                         padding = 'same', \n",
    "                         strides = strides, \n",
    "                         name = 'decoder_output')(deconv_2)\n",
    "\n",
    "decoder = Model(decoder_input, output, name = 'decoder')\n",
    "decoder.summary()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Build the VAE model\n",
    "outputs = decoder(encoder(x_input)[2])\n",
    "vae = Model(x_input, outputs, name = 'vae')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Define some losses\n",
    "def kl_loss(y_true, y_pred):\n",
    "    kl_loss = 1 + z_var - K.square(z_mean) - K.exp(z_var)\n",
    "    kl_loss = K.sum(kl_loss, axis=-1)\n",
    "    kl_loss *= -0.5\n",
    "    kl_loss = K.mean(kl_loss)\n",
    "    return kl_loss\n",
    "\n",
    "def recon_loss(y_true, y_pred):\n",
    "    recon_loss = mean_squared_error(K.flatten(y_true), K.flatten(y_pred))\n",
    "    recon_loss *= shape[0]*shape[1]*shape[2]\n",
    "    recon_loss = K.mean(recon_loss)\n",
    "    return recon_loss\n",
    "\n",
    "def my_vae_loss(y_true, y_pred):\n",
    "    recon = recon_loss(y_true, y_pred) \n",
    "    kl = kl_loss(y_true, y_pred)\n",
    "    vae_loss = recon + kl\n",
    "    return vae_loss\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"vae\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "Input (InputLayer)           (None, 4, 3, 5)           0         \n",
      "_________________________________________________________________\n",
      "encoder (Model)              [(None, 4), (None, 4), (N 19914     \n",
      "_________________________________________________________________\n",
      "decoder (Model)              (None, 4, 3, 5)           7121      \n",
      "=================================================================\n",
      "Total params: 27,035\n",
      "Trainable params: 27,035\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "vae.compile(optimizer = 'rmsprop', loss = my_vae_loss, metrics = [kl_loss, recon_loss])\n",
    "vae.summary()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 6075 samples, validate on 6075 samples\n",
      "Epoch 1/100\n",
      " - 6s - loss: 56.5557 - kl_loss: 0.6183 - recon_loss: 52.3071 - val_loss: 56.2025 - val_kl_loss: 0.5428 - val_recon_loss: 51.4786\n",
      "Epoch 2/100\n",
      " - 1s - loss: 56.1445 - kl_loss: 0.5745 - recon_loss: 50.9407 - val_loss: 56.1662 - val_kl_loss: 0.6207 - val_recon_loss: 51.3741\n",
      "Epoch 3/100\n",
      " - 2s - loss: 56.1438 - kl_loss: 0.5945 - recon_loss: 51.2637 - val_loss: 56.1303 - val_kl_loss: 0.5964 - val_recon_loss: 51.3313\n",
      "Epoch 4/100\n",
      " - 1s - loss: 56.1436 - kl_loss: 0.6347 - recon_loss: 52.7860 - val_loss: 56.1836 - val_kl_loss: 0.6258 - val_recon_loss: 51.3698\n",
      "Epoch 5/100\n",
      " - 1s - loss: 56.1807 - kl_loss: 0.6224 - recon_loss: 51.8240 - val_loss: 56.2118 - val_kl_loss: 0.6657 - val_recon_loss: 51.3802\n",
      "Epoch 6/100\n",
      " - 1s - loss: 56.1134 - kl_loss: 0.6308 - recon_loss: 53.4610 - val_loss: 56.1595 - val_kl_loss: 0.6404 - val_recon_loss: 51.3559\n",
      "Epoch 7/100\n",
      " - 1s - loss: 56.1582 - kl_loss: 0.6093 - recon_loss: 51.2590 - val_loss: 56.3265 - val_kl_loss: 0.6957 - val_recon_loss: 51.4757\n",
      "Epoch 8/100\n",
      " - 2s - loss: 56.1711 - kl_loss: 0.6085 - recon_loss: 51.0291 - val_loss: 56.1492 - val_kl_loss: 0.6240 - val_recon_loss: 51.3504\n",
      "Epoch 9/100\n",
      " - 1s - loss: 56.0964 - kl_loss: 0.6117 - recon_loss: 51.7545 - val_loss: 56.2084 - val_kl_loss: 0.6418 - val_recon_loss: 51.3699\n",
      "Epoch 10/100\n",
      " - 1s - loss: 56.0734 - kl_loss: 0.6136 - recon_loss: 51.1988 - val_loss: 56.2000 - val_kl_loss: 0.6264 - val_recon_loss: 51.3918\n",
      "Epoch 11/100\n",
      " - 1s - loss: 56.1194 - kl_loss: 0.6326 - recon_loss: 52.8405 - val_loss: 56.4416 - val_kl_loss: 0.6835 - val_recon_loss: 51.5985\n",
      "Epoch 12/100\n",
      " - 2s - loss: 56.1507 - kl_loss: 0.6345 - recon_loss: 52.9209 - val_loss: 56.3370 - val_kl_loss: 0.5724 - val_recon_loss: 51.5529\n",
      "Epoch 13/100\n",
      " - 1s - loss: 56.1603 - kl_loss: 0.6230 - recon_loss: 52.3722 - val_loss: 56.1388 - val_kl_loss: 0.6121 - val_recon_loss: 51.3364\n",
      "Epoch 14/100\n",
      " - 1s - loss: 56.0874 - kl_loss: 0.6194 - recon_loss: 51.9298 - val_loss: 56.1033 - val_kl_loss: 0.5716 - val_recon_loss: 51.3504\n",
      "Epoch 15/100\n",
      " - 1s - loss: 56.1552 - kl_loss: 0.6301 - recon_loss: 53.6996 - val_loss: 56.0579 - val_kl_loss: 0.6208 - val_recon_loss: 51.2457\n",
      "Epoch 16/100\n",
      " - 1s - loss: 56.1110 - kl_loss: 0.6257 - recon_loss: 52.6520 - val_loss: 56.1065 - val_kl_loss: 0.6107 - val_recon_loss: 51.3232\n",
      "Epoch 17/100\n",
      " - 1s - loss: 56.1233 - kl_loss: 0.6202 - recon_loss: 51.1929 - val_loss: 56.0418 - val_kl_loss: 0.6039 - val_recon_loss: 51.2743\n",
      "Epoch 18/100\n",
      " - 1s - loss: 56.0491 - kl_loss: 0.6464 - recon_loss: 62.5828 - val_loss: 56.2869 - val_kl_loss: 0.6678 - val_recon_loss: 51.4584\n",
      "Epoch 19/100\n",
      " - 1s - loss: 56.1193 - kl_loss: 0.6475 - recon_loss: 55.7429 - val_loss: 56.0833 - val_kl_loss: 0.6274 - val_recon_loss: 51.2536\n",
      "Epoch 20/100\n",
      " - 1s - loss: 56.1582 - kl_loss: 0.6323 - recon_loss: 51.3621 - val_loss: 56.1015 - val_kl_loss: 0.6529 - val_recon_loss: 51.2708\n",
      "Epoch 21/100\n",
      " - 1s - loss: 56.0219 - kl_loss: 0.6541 - recon_loss: 52.3774 - val_loss: 56.0158 - val_kl_loss: 0.6193 - val_recon_loss: 51.2115\n",
      "Epoch 22/100\n",
      " - 1s - loss: 56.0251 - kl_loss: 0.6444 - recon_loss: 57.1722 - val_loss: 56.4308 - val_kl_loss: 0.5921 - val_recon_loss: 51.6530\n",
      "Epoch 23/100\n",
      " - 0s - loss: 56.2337 - kl_loss: 0.6352 - recon_loss: 51.1531 - val_loss: 56.0675 - val_kl_loss: 0.5979 - val_recon_loss: 51.2629\n",
      "Epoch 24/100\n",
      " - 0s - loss: 56.0620 - kl_loss: 0.6253 - recon_loss: 51.3146 - val_loss: 56.2578 - val_kl_loss: 0.6899 - val_recon_loss: 51.3661\n",
      "Epoch 25/100\n",
      " - 0s - loss: 56.1223 - kl_loss: 0.6594 - recon_loss: 54.0028 - val_loss: 56.0957 - val_kl_loss: 0.6266 - val_recon_loss: 51.2768\n",
      "Epoch 26/100\n",
      " - 0s - loss: 56.0557 - kl_loss: 0.6492 - recon_loss: 57.7373 - val_loss: 56.1581 - val_kl_loss: 0.6932 - val_recon_loss: 51.2842\n",
      "Epoch 27/100\n",
      " - 0s - loss: 56.1013 - kl_loss: 0.6551 - recon_loss: 51.8052 - val_loss: 56.0969 - val_kl_loss: 0.6051 - val_recon_loss: 51.3120\n",
      "Epoch 28/100\n",
      " - 0s - loss: 56.0855 - kl_loss: 0.6491 - recon_loss: 52.2508 - val_loss: 56.6607 - val_kl_loss: 0.7542 - val_recon_loss: 51.7141\n",
      "Epoch 29/100\n",
      " - 1s - loss: 56.1709 - kl_loss: 0.6474 - recon_loss: 51.0228 - val_loss: 56.0887 - val_kl_loss: 0.6401 - val_recon_loss: 51.2528\n",
      "Epoch 30/100\n",
      " - 1s - loss: 56.0223 - kl_loss: 0.6254 - recon_loss: 54.3671 - val_loss: 56.4286 - val_kl_loss: 0.6977 - val_recon_loss: 51.5460\n",
      "Epoch 31/100\n",
      " - 1s - loss: 56.1226 - kl_loss: 0.6321 - recon_loss: 51.6298 - val_loss: 56.0838 - val_kl_loss: 0.6527 - val_recon_loss: 51.2581\n",
      "Epoch 32/100\n",
      " - 1s - loss: 56.1048 - kl_loss: 0.6494 - recon_loss: 53.8494 - val_loss: 56.1198 - val_kl_loss: 0.6276 - val_recon_loss: 51.2900\n",
      "Epoch 33/100\n",
      " - 1s - loss: 56.0556 - kl_loss: 0.6345 - recon_loss: 51.6564 - val_loss: 56.1385 - val_kl_loss: 0.6525 - val_recon_loss: 51.2988\n",
      "Epoch 34/100\n",
      " - 1s - loss: 56.0672 - kl_loss: 0.6549 - recon_loss: 51.9880 - val_loss: 56.0511 - val_kl_loss: 0.6402 - val_recon_loss: 51.2417\n",
      "Epoch 35/100\n",
      " - 0s - loss: 56.0071 - kl_loss: 0.7006 - recon_loss: 80.2109 - val_loss: 56.1704 - val_kl_loss: 0.5912 - val_recon_loss: 51.4248\n",
      "Epoch 36/100\n",
      " - 1s - loss: 56.0443 - kl_loss: 0.6416 - recon_loss: 53.5215 - val_loss: 56.0420 - val_kl_loss: 0.6600 - val_recon_loss: 51.2099\n",
      "Epoch 37/100\n",
      " - 1s - loss: 56.0743 - kl_loss: 0.6465 - recon_loss: 51.5680 - val_loss: 56.0332 - val_kl_loss: 0.6704 - val_recon_loss: 51.1664\n",
      "Epoch 38/100\n",
      " - 1s - loss: 56.0099 - kl_loss: 0.6505 - recon_loss: 51.9589 - val_loss: 56.0153 - val_kl_loss: 0.6585 - val_recon_loss: 51.1896\n",
      "Epoch 39/100\n",
      " - 1s - loss: 55.9658 - kl_loss: 0.6330 - recon_loss: 51.0514 - val_loss: 56.0455 - val_kl_loss: 0.6397 - val_recon_loss: 51.2167\n",
      "Epoch 40/100\n",
      " - 1s - loss: 56.1457 - kl_loss: 0.6440 - recon_loss: 51.3686 - val_loss: 56.0635 - val_kl_loss: 0.6617 - val_recon_loss: 51.2150\n",
      "Epoch 41/100\n",
      " - 1s - loss: 55.9867 - kl_loss: 0.6605 - recon_loss: 61.8846 - val_loss: 55.9837 - val_kl_loss: 0.6521 - val_recon_loss: 51.1259\n",
      "Epoch 42/100\n",
      " - 1s - loss: 55.9633 - kl_loss: 0.6516 - recon_loss: 52.6736 - val_loss: 56.1673 - val_kl_loss: 0.6967 - val_recon_loss: 51.2430\n",
      "Epoch 43/100\n",
      " - 1s - loss: 56.0963 - kl_loss: 0.6510 - recon_loss: 51.5498 - val_loss: 56.0191 - val_kl_loss: 0.6411 - val_recon_loss: 51.1766\n",
      "Epoch 44/100\n",
      " - 0s - loss: 56.0139 - kl_loss: 0.6451 - recon_loss: 52.7982 - val_loss: 56.0548 - val_kl_loss: 0.6375 - val_recon_loss: 51.2172\n",
      "Epoch 45/100\n",
      " - 0s - loss: 55.9696 - kl_loss: 0.6410 - recon_loss: 51.3871 - val_loss: 55.9696 - val_kl_loss: 0.6407 - val_recon_loss: 51.1438\n",
      "Epoch 46/100\n",
      " - 0s - loss: 56.0302 - kl_loss: 0.6902 - recon_loss: 70.3234 - val_loss: 56.4691 - val_kl_loss: 0.7629 - val_recon_loss: 51.5097\n",
      "Epoch 47/100\n",
      " - 1s - loss: 56.0487 - kl_loss: 0.6700 - recon_loss: 50.8327 - val_loss: 55.9500 - val_kl_loss: 0.6496 - val_recon_loss: 51.0731\n",
      "Epoch 48/100\n",
      " - 0s - loss: 55.9514 - kl_loss: 0.6249 - recon_loss: 51.0818 - val_loss: 56.1573 - val_kl_loss: 0.6720 - val_recon_loss: 51.3144\n",
      "Epoch 49/100\n",
      " - 1s - loss: 56.0065 - kl_loss: 0.6622 - recon_loss: 52.4647 - val_loss: 56.0700 - val_kl_loss: 0.6444 - val_recon_loss: 51.2325\n",
      "Epoch 50/100\n",
      " - 1s - loss: 56.0237 - kl_loss: 0.6539 - recon_loss: 51.0289 - val_loss: 56.1927 - val_kl_loss: 0.7148 - val_recon_loss: 51.2676\n",
      "Epoch 51/100\n",
      " - 1s - loss: 56.0563 - kl_loss: 0.6546 - recon_loss: 51.3513 - val_loss: 56.0429 - val_kl_loss: 0.6377 - val_recon_loss: 51.1775\n",
      "Epoch 52/100\n",
      " - 1s - loss: 55.9734 - kl_loss: 0.6683 - recon_loss: 51.6604 - val_loss: 56.1195 - val_kl_loss: 0.7116 - val_recon_loss: 51.2351\n",
      "Epoch 53/100\n",
      " - 1s - loss: 56.0488 - kl_loss: 0.6681 - recon_loss: 51.9234 - val_loss: 55.9807 - val_kl_loss: 0.6850 - val_recon_loss: 51.0817\n",
      "Epoch 54/100\n",
      " - 0s - loss: 55.9614 - kl_loss: 0.6701 - recon_loss: 55.4946 - val_loss: 56.2318 - val_kl_loss: 0.7340 - val_recon_loss: 51.2930\n",
      "Epoch 55/100\n",
      " - 0s - loss: 56.0416 - kl_loss: 0.6606 - recon_loss: 51.4948 - val_loss: 55.9440 - val_kl_loss: 0.6502 - val_recon_loss: 51.0831\n",
      "Epoch 56/100\n",
      " - 1s - loss: 55.9382 - kl_loss: 0.6831 - recon_loss: 81.5232 - val_loss: 56.0712 - val_kl_loss: 0.6440 - val_recon_loss: 51.2543\n",
      "Epoch 57/100\n",
      " - 1s - loss: 56.0798 - kl_loss: 0.6691 - recon_loss: 52.4877 - val_loss: 56.2543 - val_kl_loss: 0.6928 - val_recon_loss: 51.3495\n",
      "Epoch 58/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 1s - loss: 56.0608 - kl_loss: 0.6694 - recon_loss: 51.4153 - val_loss: 56.1559 - val_kl_loss: 0.6960 - val_recon_loss: 51.2753\n",
      "Epoch 59/100\n",
      " - 1s - loss: 55.9875 - kl_loss: 0.6876 - recon_loss: 51.5730 - val_loss: 56.0438 - val_kl_loss: 0.6551 - val_recon_loss: 51.1946\n",
      "Epoch 60/100\n",
      " - 1s - loss: 55.9682 - kl_loss: 0.6789 - recon_loss: 54.0749 - val_loss: 56.0118 - val_kl_loss: 0.6208 - val_recon_loss: 51.1669\n",
      "Epoch 61/100\n",
      " - 1s - loss: 55.9752 - kl_loss: 0.6570 - recon_loss: 53.5645 - val_loss: 56.0601 - val_kl_loss: 0.6585 - val_recon_loss: 51.1895\n",
      "Epoch 62/100\n",
      " - 1s - loss: 55.9949 - kl_loss: 0.6749 - recon_loss: 51.0826 - val_loss: 56.1387 - val_kl_loss: 0.6325 - val_recon_loss: 51.2855\n",
      "Epoch 63/100\n",
      " - 1s - loss: 55.9716 - kl_loss: 0.6473 - recon_loss: 51.1262 - val_loss: 55.9852 - val_kl_loss: 0.6904 - val_recon_loss: 51.0847\n",
      "Epoch 64/100\n",
      " - 1s - loss: 56.0319 - kl_loss: 0.6801 - recon_loss: 50.9433 - val_loss: 56.0185 - val_kl_loss: 0.6740 - val_recon_loss: 51.1568\n",
      "Epoch 65/100\n",
      " - 1s - loss: 55.9226 - kl_loss: 0.6820 - recon_loss: 52.8010 - val_loss: 56.0731 - val_kl_loss: 0.7100 - val_recon_loss: 51.1791\n",
      "Epoch 66/100\n",
      " - 1s - loss: 55.9727 - kl_loss: 0.6916 - recon_loss: 52.5401 - val_loss: 55.9860 - val_kl_loss: 0.6557 - val_recon_loss: 51.1309\n",
      "Epoch 67/100\n",
      " - 1s - loss: 55.9264 - kl_loss: 0.6485 - recon_loss: 51.4103 - val_loss: 56.4218 - val_kl_loss: 0.7776 - val_recon_loss: 51.4698\n",
      "Epoch 68/100\n",
      " - 1s - loss: 56.0380 - kl_loss: 0.6747 - recon_loss: 51.1272 - val_loss: 55.9274 - val_kl_loss: 0.6460 - val_recon_loss: 51.0736\n",
      "Epoch 69/100\n",
      " - 1s - loss: 55.9203 - kl_loss: 0.6558 - recon_loss: 51.4768 - val_loss: 56.0032 - val_kl_loss: 0.6850 - val_recon_loss: 51.1299\n",
      "Epoch 70/100\n",
      " - 0s - loss: 55.9083 - kl_loss: 0.6641 - recon_loss: 51.0350 - val_loss: 55.9018 - val_kl_loss: 0.6998 - val_recon_loss: 51.0011\n",
      "Epoch 71/100\n",
      " - 0s - loss: 56.0141 - kl_loss: 0.6841 - recon_loss: 50.7147 - val_loss: 55.9434 - val_kl_loss: 0.7171 - val_recon_loss: 51.0128\n",
      "Epoch 72/100\n",
      " - 0s - loss: 55.9235 - kl_loss: 0.6797 - recon_loss: 50.6320 - val_loss: 56.0023 - val_kl_loss: 0.6596 - val_recon_loss: 51.1338\n",
      "Epoch 73/100\n",
      " - 1s - loss: 55.9649 - kl_loss: 0.6647 - recon_loss: 51.1179 - val_loss: 55.9317 - val_kl_loss: 0.7018 - val_recon_loss: 51.0215\n",
      "Epoch 74/100\n",
      " - 0s - loss: 55.9117 - kl_loss: 0.6984 - recon_loss: 60.6602 - val_loss: 55.9970 - val_kl_loss: 0.6591 - val_recon_loss: 51.1669\n",
      "Epoch 75/100\n",
      " - 0s - loss: 55.9024 - kl_loss: 0.6710 - recon_loss: 53.0012 - val_loss: 56.4160 - val_kl_loss: 0.7412 - val_recon_loss: 51.4536\n",
      "Epoch 76/100\n",
      " - 0s - loss: 56.0567 - kl_loss: 0.6848 - recon_loss: 51.4966 - val_loss: 55.9350 - val_kl_loss: 0.7003 - val_recon_loss: 51.0225\n",
      "Epoch 77/100\n",
      " - 1s - loss: 56.0005 - kl_loss: 0.7397 - recon_loss: 142.8502 - val_loss: 56.1652 - val_kl_loss: 0.6289 - val_recon_loss: 51.3332\n",
      "Epoch 78/100\n",
      " - 1s - loss: 55.9564 - kl_loss: 0.6546 - recon_loss: 51.2339 - val_loss: 55.8802 - val_kl_loss: 0.6792 - val_recon_loss: 50.9798\n",
      "Epoch 79/100\n",
      " - 0s - loss: 55.9051 - kl_loss: 0.6719 - recon_loss: 50.9725 - val_loss: 56.0107 - val_kl_loss: 0.6441 - val_recon_loss: 51.1852\n",
      "Epoch 80/100\n",
      " - 0s - loss: 55.9050 - kl_loss: 0.6718 - recon_loss: 51.9200 - val_loss: 55.9714 - val_kl_loss: 0.6462 - val_recon_loss: 51.1078\n",
      "Epoch 81/100\n",
      " - 0s - loss: 55.9254 - kl_loss: 0.6668 - recon_loss: 51.1212 - val_loss: 56.0166 - val_kl_loss: 0.6718 - val_recon_loss: 51.1421\n",
      "Epoch 82/100\n",
      " - 1s - loss: 55.9962 - kl_loss: 0.6926 - recon_loss: 55.2406 - val_loss: 56.3537 - val_kl_loss: 0.7263 - val_recon_loss: 51.4214\n",
      "Epoch 83/100\n",
      " - 1s - loss: 55.9951 - kl_loss: 0.7003 - recon_loss: 67.7276 - val_loss: 55.9497 - val_kl_loss: 0.6742 - val_recon_loss: 51.1017\n",
      "Epoch 84/100\n",
      " - 1s - loss: 55.8812 - kl_loss: 0.6796 - recon_loss: 52.7741 - val_loss: 56.0316 - val_kl_loss: 0.6727 - val_recon_loss: 51.1203\n",
      "Epoch 85/100\n",
      " - 1s - loss: 55.9538 - kl_loss: 0.6745 - recon_loss: 51.0025 - val_loss: 56.0843 - val_kl_loss: 0.7334 - val_recon_loss: 51.1240\n",
      "Epoch 86/100\n",
      " - 1s - loss: 55.9619 - kl_loss: 0.6934 - recon_loss: 51.0403 - val_loss: 55.9875 - val_kl_loss: 0.6505 - val_recon_loss: 51.1556\n",
      "Epoch 87/100\n",
      " - 1s - loss: 55.9518 - kl_loss: 0.6699 - recon_loss: 50.9785 - val_loss: 55.9607 - val_kl_loss: 0.6278 - val_recon_loss: 51.1351\n",
      "Epoch 88/100\n",
      " - 0s - loss: 55.8851 - kl_loss: 0.6511 - recon_loss: 51.4951 - val_loss: 56.1919 - val_kl_loss: 0.6900 - val_recon_loss: 51.3012\n",
      "Epoch 89/100\n",
      " - 0s - loss: 56.0311 - kl_loss: 0.6899 - recon_loss: 52.7530 - val_loss: 55.9346 - val_kl_loss: 0.7060 - val_recon_loss: 51.0285\n",
      "Epoch 90/100\n",
      " - 0s - loss: 55.8667 - kl_loss: 0.6762 - recon_loss: 51.9401 - val_loss: 56.2255 - val_kl_loss: 0.6518 - val_recon_loss: 51.3772\n",
      "Epoch 91/100\n",
      " - 1s - loss: 55.9602 - kl_loss: 0.7014 - recon_loss: 51.0049 - val_loss: 55.9028 - val_kl_loss: 0.7521 - val_recon_loss: 50.9481\n",
      "Epoch 92/100\n",
      " - 1s - loss: 55.9093 - kl_loss: 0.7178 - recon_loss: 66.3444 - val_loss: 55.9876 - val_kl_loss: 0.6454 - val_recon_loss: 51.1024\n",
      "Epoch 93/100\n",
      " - 1s - loss: 55.9516 - kl_loss: 0.6629 - recon_loss: 51.2175 - val_loss: 55.9939 - val_kl_loss: 0.6463 - val_recon_loss: 51.1662\n",
      "Epoch 94/100\n",
      " - 1s - loss: 55.8627 - kl_loss: 0.6817 - recon_loss: 51.9826 - val_loss: 55.8941 - val_kl_loss: 0.6804 - val_recon_loss: 51.0094\n",
      "Epoch 95/100\n",
      " - 1s - loss: 55.8372 - kl_loss: 0.6816 - recon_loss: 101.2709 - val_loss: 56.0584 - val_kl_loss: 0.7510 - val_recon_loss: 51.1062\n",
      "Epoch 96/100\n",
      " - 1s - loss: 56.0051 - kl_loss: 0.6956 - recon_loss: 51.9095 - val_loss: 55.8882 - val_kl_loss: 0.7198 - val_recon_loss: 50.9547\n",
      "Epoch 97/100\n",
      " - 1s - loss: 55.8621 - kl_loss: 0.6903 - recon_loss: 51.6162 - val_loss: 55.9515 - val_kl_loss: 0.6612 - val_recon_loss: 51.0804\n",
      "Epoch 98/100\n",
      " - 1s - loss: 55.9042 - kl_loss: 0.6718 - recon_loss: 50.7287 - val_loss: 55.9776 - val_kl_loss: 0.7073 - val_recon_loss: 51.0407\n",
      "Epoch 99/100\n",
      " - 1s - loss: 55.9225 - kl_loss: 0.6917 - recon_loss: 51.4020 - val_loss: 55.8813 - val_kl_loss: 0.7185 - val_recon_loss: 50.9445\n",
      "Epoch 100/100\n",
      " - 1s - loss: 55.8392 - kl_loss: 0.6918 - recon_loss: 51.8695 - val_loss: 55.9692 - val_kl_loss: 0.6573 - val_recon_loss: 51.1405\n"
     ]
    }
   ],
   "source": [
    "#Fit the model\n",
    "epochs = 100\n",
    "batch_size = 1000\n",
    "\n",
    "history = vae.fit(x = x_train,\n",
    "                  y = x_train,\n",
    "                  validation_data = (x_train, x_train),\n",
    "                  epochs = epochs,\n",
    "                  batch_size = batch_size,\n",
    "                  verbose = 2)\n",
    "\n",
    "vae.save('cnn-vae_model.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "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": [
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('Model loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['loss', 'kl_loss', 'recon_loss']"
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vae.metrics_names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "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": [
    "plt.plot(history.history['kl_loss'])\n",
    "plt.title('Kullback-Liebler Loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc = 'upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "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": [
    "plt.plot(history.history['recon_loss'])\n",
    "plt.title('Reconstruction (MSE) Loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc = 'upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "anom_df = csv_to_df('bsm_test/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "anom_df = anom_df.fillna(0)\n",
    "anom_vars = [entry for entry in df.columns if entry[0] == 'E'] + [entry for entry in df.columns if entry[0:2] == 'pt'] + [entry for entry in df.columns if entry[0:2] == 'et'] + [entry for entry in df.columns if entry[0:2] == 'ph']\n",
    "#Now we one_hot encode the data to make sure we have a class definition\n",
    "anom_df = anom_df[['process_ID']+anom_vars]\n",
    "anom_one_hot = pd.get_dummies(anom_df['process_ID'])\n",
    "anom_processes = anom_one_hot.columns\n",
    "anom_df.drop('process_ID', axis = 'columns', inplace = True)\n",
    "anom_df = pd.concat([anom_df, anom_one_hot], sort = False, axis = 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [],
   "source": [
    "#We have created a df of our values and some kind of class label\n",
    "#We should normalize\n",
    "anom_x = anom_df[anom_vars].values\n",
    "anom_scaled = StandardScaler().fit_transform(anom_x)\n",
    "anom_df[anom_vars] = anom_scaled\n",
    "\n",
    "anom_x_train, anom_x_test, anom_y_train, anom_y_test = train_test_split(anom_df[anom_vars].values,\n",
    "\t\t\t\t\t\t    anom_df[anom_processes].values, \n",
    "                                                    shuffle = True,\n",
    "                                                    random_state = 42,\n",
    "                                                    test_size = 0.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Reshape the data. This will have to be done explicitly until you think of a clever way of doing it\n",
    "anom_x_train = anom_x_train.reshape((anom_x_train.shape[0], 4, 3, 5))\n",
    "anom_x_test = anom_x_test.reshape((anom_x_test.shape[0], 4, 3, 5))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import load_model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2455/2455 [==============================] - 0s 149us/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "62.426984195748076"
      ]
     },
     "execution_count": 93,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vae.evaluate(x_train[:2455], anom_x_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}