autoencoder/anomaly-detection-training-autoencoder.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "from os import listdir\n",
    "from os.path import join\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import random\n",
    "from scipy import stats\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras import layers, models, optimizers, regularizers\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import seaborn as sn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Python 3.7.6\n",
      "Numpy 1.18.1\n",
      "TensorFlow 2.1.0\n",
      "Keras 2.2.4-tf\n"
     ]
    }
   ],
   "source": [
    "# Print versions\n",
    "!python --version\n",
    "print('Numpy ' + np.__version__)\n",
    "print('TensorFlow ' + tf.__version__)\n",
    "print('Keras ' + tf.keras.__version__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Max measurements per file: 128\n",
      "Downsample factor: 2\n",
      "Window length: 64\n"
     ]
    }
   ],
   "source": [
    "# Settings\n",
    "dataset_path = '..\\\\datasets\\\\ceiling-fan'  # Directory where raw accelerometer data is stored\n",
    "normal_op_list = ['fan_0_low_0_weight-moving'] #['fan_0_low_0_weight']\n",
    "anomaly_op_list = ['fan_0_med_0_weight', 'fan_0_high_0_weight',\n",
    "                  'fan_0_low_1_weight', 'fan_0_med_1_weight', 'fan_0_high_1_weight']\n",
    "val_ratio = 0.2             # Percentage of samples that should be held for validation set\n",
    "test_ratio = 0.2            # Percentage of samples that should be held for test set\n",
    "raw_scale = 1               # Multiply raw values to fit into integers\n",
    "sensor_sample_rate = 200    # Hz\n",
    "desired_sample_rate = 50    # Hz\n",
    "sample_time = 0.64           # Time (sec) length of each sample\n",
    "samples_per_file = 128      # Expected number of measurements in each file (truncate to this)\n",
    "max_measurements = int(sample_time * sensor_sample_rate)\n",
    "downsample_factor = int(samples_per_file / desired_sample_rate)\n",
    "win_len = int(max_measurements / downsample_factor)\n",
    "\n",
    "keras_model_name = 'models\\\\fan_low_model-moving'           # Will be given .h5 suffix\n",
    "sample_file_name = '..\\\\test_samples\\\\normal_anomaly_samples'  # Will be given .npz suffix\n",
    "rep_dataset_name = '..\\\\test_samples\\\\normal_anomaly_test_set' # Will be given .npz suffix\n",
    "\n",
    "print('Max measurements per file:', max_measurements)\n",
    "print('Downsample factor:', downsample_factor)\n",
    "print('Window length:', win_len)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create list of filenames\n",
    "def createFilenameList(op_list):\n",
    "    \n",
    "    # Extract paths and filenames in each directory\n",
    "    op_filenames = []\n",
    "    num_samples = 0\n",
    "    for index, target in enumerate(op_list):\n",
    "        samples_in_dir = listdir(join(dataset_path, target))\n",
    "        samples_in_dir = [join(dataset_path, target, sample) for sample in samples_in_dir]\n",
    "        op_filenames.append(samples_in_dir)\n",
    "    \n",
    "    # Flatten list\n",
    "    return [item for sublist in op_filenames for item in sublist]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of normal samples: 299\n",
      "Number of anomaly samples: 2926\n"
     ]
    }
   ],
   "source": [
    "# Create normal and anomaly filename lists\n",
    "normal_op_filenames = createFilenameList(normal_op_list)\n",
    "anomaly_op_filenames = createFilenameList(anomaly_op_list)\n",
    "print('Number of normal samples:', len(normal_op_filenames))\n",
    "print('Number of anomaly samples:', len(anomaly_op_filenames))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Shuffle lists\n",
    "random.shuffle(normal_op_filenames)\n",
    "random.shuffle(anomaly_op_filenames)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Calculate validation and test set sizes\n",
    "val_set_size = int(len(normal_op_filenames) * val_ratio)\n",
    "test_set_size = int(len(normal_op_filenames) * test_ratio)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of training samples: 181\n",
      "Number of validation samples: 59\n",
      "Number of test samples: 59\n"
     ]
    }
   ],
   "source": [
    "# Break dataset apart into train, validation, and test sets\n",
    "num_samples = len(normal_op_filenames)\n",
    "filenames_val = normal_op_filenames[:val_set_size]\n",
    "filenames_test = normal_op_filenames[val_set_size:(val_set_size + test_set_size)]\n",
    "filenames_train = normal_op_filenames[(val_set_size + test_set_size):]\n",
    "\n",
    "# Print out number of samples in each set\n",
    "print('Number of training samples:', len(filenames_train))\n",
    "print('Number of validation samples:', len(filenames_val))\n",
    "print('Number of test samples:', len(filenames_test))\n",
    "\n",
    "# Check that our splits add up correctly\n",
    "assert(len(filenames_train) + len(filenames_val) + len(filenames_test)) == num_samples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Function: extract specified features (variances, MAD) from sample\n",
    "def extract_features(sample, max_measurements=0, scale=1):\n",
    "    \n",
    "    features = []\n",
    "    \n",
    "    # Truncate sample\n",
    "    if max_measurements == 0:\n",
    "        max_measurements = sample.shape[0]\n",
    "    sample = sample[0:max_measurements]\n",
    "    \n",
    "    # Scale sample\n",
    "    sample = scale * sample\n",
    "    \n",
    "        \n",
    "#     # Remove DC component\n",
    "#     sample = sample - np.mean(sample, axis=0)\n",
    "    \n",
    "#     # Truncate sample\n",
    "#     sample = sample[0:max_measurements]\n",
    "    \n",
    "#     # Variance\n",
    "#     features.append(np.var(sample, axis=0))\n",
    "    \n",
    "#     # Kurtosis\n",
    "#     features.append(stats.kurtosis(sample))\n",
    "    \n",
    "#     # Skew\n",
    "#     features.append(stats.skew(sample))\n",
    "    \n",
    "    # Median absolute deviation (MAD)\n",
    "    features.append(stats.median_absolute_deviation(sample))\n",
    "    \n",
    "#     # Correlation\n",
    "#     cov = np.corrcoef(sample.T)\n",
    "#     features.append(np.array([cov[0,1], cov[0,2], cov[1,2]]))\n",
    "    \n",
    "    # Compute a windowed FFT of each axis in the sample (leave off DC)\n",
    "#     sample = sample[::downsample_factor, :]  # Downsample\n",
    "#     sample = np.floor(sample)                # Round down to int\n",
    "#     hann_window = np.hanning(sample.shape[0])\n",
    "#     for i, axis in enumerate(sample.T):\n",
    "#         fft = abs(np.fft.rfft(axis * hann_window))\n",
    "#         features.append(fft[1:])  # Leave off DC\n",
    "    \n",
    "    return np.array(features).flatten()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(3,)\n",
      "[0.00615279 0.00217201 0.00398226]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x154ba56c488>]"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "# Test with 1 sample\n",
    "sample = np.genfromtxt(filenames_test[0], delimiter=',')\n",
    "features = extract_features(sample, max_measurements, scale=raw_scale)\n",
    "print(features.shape)\n",
    "print(features)\n",
    "plt.plot(features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Function: loop through filenames, creating feature sets\n",
    "def create_feature_set(filenames):\n",
    "    x_out = []\n",
    "    for file in filenames:\n",
    "        sample = np.genfromtxt(file, delimiter=',')\n",
    "        features = extract_features(sample, max_measurements, raw_scale)\n",
    "        x_out.append(features)\n",
    "        \n",
    "    return np.array(x_out)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracted features from training set. Shape: (181, 3)\n",
      "Extracted features from validation set. Shape: (59, 3)\n",
      "Extracted features from test set. Shape: (59, 3)\n"
     ]
    }
   ],
   "source": [
    "# Create training, validation, and test sets\n",
    "x_train = create_feature_set(filenames_train)\n",
    "print('Extracted features from training set. Shape:', x_train.shape)\n",
    "x_val = create_feature_set(filenames_val)\n",
    "print('Extracted features from validation set. Shape:', x_val.shape)\n",
    "x_test = create_feature_set(filenames_test)\n",
    "print('Extracted features from test set. Shape:', x_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(3,)\n"
     ]
    }
   ],
   "source": [
    "# Get input shape for 1 sample\n",
    "sample_shape = x_train.shape[1:]\n",
    "print(sample_shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_10\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_20 (Dense)             (None, 2)                 8         \n",
      "_________________________________________________________________\n",
      "dropout_10 (Dropout)         (None, 2)                 0         \n",
      "_________________________________________________________________\n",
      "dense_21 (Dense)             (None, 3)                 9         \n",
      "=================================================================\n",
      "Total params: 17\n",
      "Trainable params: 17\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Build model\n",
    "# Based on: https://medium.com/@curiousily/credit-card-fraud-detection-using-autoencoders-in-keras-tensorflow-for-hackers-part-vii-20e0c85301bd\n",
    "encoding_dim = 2       # Number of nodes in first layer\n",
    "model = models.Sequential([\n",
    "    layers.InputLayer(input_shape=sample_shape),\n",
    "    layers.Dense(encoding_dim, activation='relu'),\n",
    "    layers.Dropout(0.2),\n",
    "    layers.Dense(*sample_shape, activation='relu')\n",
    "])\n",
    "\n",
    "# Display model\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Add training parameters to model\n",
    "model.compile(optimizer='adam',\n",
    "             loss='mse')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 181 samples, validate on 59 samples\n",
      "Epoch 1/50\n",
      "181/181 [==============================] - 0s 2ms/sample - loss: 4.1740e-05 - val_loss: 2.7205e-05\n",
      "Epoch 2/50\n",
      "181/181 [==============================] - 0s 55us/sample - loss: 3.6577e-05 - val_loss: 3.0068e-05\n",
      "Epoch 3/50\n",
      "181/181 [==============================] - 0s 66us/sample - loss: 3.5605e-05 - val_loss: 2.9647e-05\n",
      "Epoch 4/50\n",
      "181/181 [==============================] - 0s 61us/sample - loss: 3.6264e-05 - val_loss: 2.3591e-05\n",
      "Epoch 5/50\n",
      "181/181 [==============================] - 0s 66us/sample - loss: 2.9814e-05 - val_loss: 1.9860e-05\n",
      "Epoch 6/50\n",
      "181/181 [==============================] - 0s 72us/sample - loss: 2.6450e-05 - val_loss: 1.9045e-05\n",
      "Epoch 7/50\n",
      "181/181 [==============================] - 0s 55us/sample - loss: 2.6769e-05 - val_loss: 1.9133e-05\n",
      "Epoch 8/50\n",
      "181/181 [==============================] - 0s 61us/sample - loss: 2.7531e-05 - val_loss: 1.8300e-05\n",
      "Epoch 9/50\n",
      "181/181 [==============================] - 0s 61us/sample - loss: 2.6435e-05 - val_loss: 1.7111e-05\n",
      "Epoch 10/50\n",
      "181/181 [==============================] - 0s 55us/sample - loss: 2.6059e-05 - val_loss: 1.6843e-05\n",
      "Epoch 11/50\n",
      "181/181 [==============================] - 0s 55us/sample - loss: 2.4881e-05 - val_loss: 1.7164e-05\n",
      "Epoch 12/50\n",
      "181/181 [==============================] - 0s 50us/sample - loss: 2.4534e-05 - val_loss: 1.7360e-05\n",
      "Epoch 13/50\n",
      "181/181 [==============================] - 0s 55us/sample - loss: 2.4759e-05 - val_loss: 1.7041e-05\n",
      "Epoch 14/50\n",
      "181/181 [==============================] - 0s 72us/sample - loss: 2.3485e-05 - val_loss: 1.6409e-05\n",
      "Epoch 15/50\n",
      "181/181 [==============================] - 0s 55us/sample - loss: 2.3852e-05 - val_loss: 1.5329e-05\n",
      "Epoch 16/50\n",
      "181/181 [==============================] - 0s 66us/sample - loss: 2.3627e-05 - val_loss: 1.4035e-05\n",
      "Epoch 17/50\n",
      "181/181 [==============================] - 0s 66us/sample - loss: 2.2932e-05 - val_loss: 1.2928e-05\n",
      "Epoch 18/50\n",
      "181/181 [==============================] - 0s 61us/sample - loss: 2.2829e-05 - val_loss: 1.2387e-05\n",
      "Epoch 19/50\n",
      "181/181 [==============================] - 0s 61us/sample - loss: 2.2194e-05 - val_loss: 1.2358e-05\n",
      "Epoch 20/50\n",
      "181/181 [==============================] - 0s 61us/sample - loss: 2.1149e-05 - val_loss: 1.2578e-05\n",
      "Epoch 21/50\n",
      "181/181 [==============================] - 0s 72us/sample - loss: 1.9909e-05 - val_loss: 1.2651e-05\n",
      "Epoch 22/50\n",
      "181/181 [==============================] - 0s 50us/sample - loss: 2.0772e-05 - val_loss: 1.2398e-05\n",
      "Epoch 23/50\n",
      "181/181 [==============================] - 0s 61us/sample - loss: 1.9717e-05 - val_loss: 1.1868e-05\n",
      "Epoch 24/50\n",
      "181/181 [==============================] - 0s 66us/sample - loss: 1.9665e-05 - val_loss: 1.1298e-05\n",
      "Epoch 25/50\n",
      "181/181 [==============================] - 0s 55us/sample - loss: 1.8762e-05 - val_loss: 1.0990e-05\n",
      "Epoch 26/50\n",
      "181/181 [==============================] - 0s 66us/sample - loss: 1.8498e-05 - val_loss: 1.0823e-05\n",
      "Epoch 27/50\n",
      "181/181 [==============================] - 0s 55us/sample - loss: 1.8547e-05 - val_loss: 1.0931e-05\n",
      "Epoch 28/50\n",
      "181/181 [==============================] - 0s 77us/sample - loss: 1.8897e-05 - val_loss: 1.1073e-05\n",
      "Epoch 29/50\n",
      "181/181 [==============================] - 0s 66us/sample - loss: 1.9005e-05 - val_loss: 1.1168e-05\n",
      "Epoch 30/50\n",
      "181/181 [==============================] - 0s 55us/sample - loss: 1.8936e-05 - val_loss: 1.1163e-05\n",
      "Epoch 31/50\n",
      "181/181 [==============================] - 0s 72us/sample - loss: 1.9001e-05 - val_loss: 1.1251e-05\n",
      "Epoch 32/50\n",
      "181/181 [==============================] - 0s 61us/sample - loss: 1.8277e-05 - val_loss: 1.1107e-05\n",
      "Epoch 33/50\n",
      "181/181 [==============================] - 0s 61us/sample - loss: 1.8405e-05 - val_loss: 1.0915e-05\n",
      "Epoch 34/50\n",
      "181/181 [==============================] - 0s 66us/sample - loss: 1.8470e-05 - val_loss: 1.0698e-05\n",
      "Epoch 35/50\n",
      "181/181 [==============================] - 0s 61us/sample - loss: 1.8463e-05 - val_loss: 1.0572e-05\n",
      "Epoch 36/50\n",
      "181/181 [==============================] - 0s 83us/sample - loss: 1.8612e-05 - val_loss: 1.0631e-05\n",
      "Epoch 37/50\n",
      "181/181 [==============================] - 0s 66us/sample - loss: 1.7745e-05 - val_loss: 1.0760e-05\n",
      "Epoch 38/50\n",
      "181/181 [==============================] - 0s 72us/sample - loss: 1.8700e-05 - val_loss: 1.0810e-05\n",
      "Epoch 39/50\n",
      "181/181 [==============================] - 0s 50us/sample - loss: 1.8361e-05 - val_loss: 1.0804e-05\n",
      "Epoch 40/50\n",
      "181/181 [==============================] - 0s 61us/sample - loss: 1.8115e-05 - val_loss: 1.0642e-05\n",
      "Epoch 41/50\n",
      "181/181 [==============================] - 0s 61us/sample - loss: 1.8655e-05 - val_loss: 1.0621e-05\n",
      "Epoch 42/50\n",
      "181/181 [==============================] - 0s 61us/sample - loss: 1.7969e-05 - val_loss: 1.0556e-05\n",
      "Epoch 43/50\n",
      "181/181 [==============================] - 0s 61us/sample - loss: 1.8055e-05 - val_loss: 1.0563e-05\n",
      "Epoch 44/50\n",
      "181/181 [==============================] - 0s 55us/sample - loss: 1.8154e-05 - val_loss: 1.0585e-05\n",
      "Epoch 45/50\n",
      "181/181 [==============================] - 0s 83us/sample - loss: 1.7518e-05 - val_loss: 1.0548e-05\n",
      "Epoch 46/50\n",
      "181/181 [==============================] - 0s 72us/sample - loss: 1.8149e-05 - val_loss: 1.0431e-05\n",
      "Epoch 47/50\n",
      "181/181 [==============================] - 0s 61us/sample - loss: 1.8104e-05 - val_loss: 1.0423e-05\n",
      "Epoch 48/50\n",
      "181/181 [==============================] - 0s 72us/sample - loss: 1.8195e-05 - val_loss: 1.0547e-05\n",
      "Epoch 49/50\n",
      "181/181 [==============================] - 0s 61us/sample - loss: 1.8077e-05 - val_loss: 1.0812e-05\n",
      "Epoch 50/50\n",
      "181/181 [==============================] - 0s 72us/sample - loss: 1.7572e-05 - val_loss: 1.0876e-05\n"
     ]
    }
   ],
   "source": [
    "# Train model (note Y labels are same as inputs, X)\n",
    "history = model.fit(x_train,\n",
    "                   x_train,\n",
    "                   epochs=50,\n",
    "                   batch_size=100,\n",
    "                   validation_data=(x_val, x_val),\n",
    "                   verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "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": [
    "# Plot results\n",
    "loss = history.history['loss']\n",
    "val_loss = history.history['val_loss']\n",
    "\n",
    "epochs = range(1, len(loss) + 1)\n",
    "\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average MSE for normal validation set: 1.0876377078421948e-05\n",
      "Standard deviation of MSE for normal validation set: 9.43928085720735e-06\n",
      "Recommended threshold (3x std dev + avg): 3.9194219650043994e-05\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(array([19., 15.,  5.,  7.,  4.,  2.,  0.,  3.,  0.,  1.,  0.,  0.,  1.,\n",
       "         0.,  0.,  0.,  0.,  1.,  0.,  1.]),\n",
       " array([3.34983955e-06, 5.83323533e-06, 8.31663112e-06, 1.08000269e-05,\n",
       "        1.32834227e-05, 1.57668185e-05, 1.82502143e-05, 2.07336101e-05,\n",
       "        2.32170059e-05, 2.57004016e-05, 2.81837974e-05, 3.06671932e-05,\n",
       "        3.31505890e-05, 3.56339848e-05, 3.81173806e-05, 4.06007764e-05,\n",
       "        4.30841722e-05, 4.55675680e-05, 4.80509637e-05, 5.05343595e-05,\n",
       "        5.30177553e-05]),\n",
       " <a list of 20 Patch objects>)"
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "# Calculate MSE from validation set\n",
    "predictions = model.predict(x_val)\n",
    "normal_mse = np.mean(np.power(x_val - predictions, 2), axis=1)\n",
    "print('Average MSE for normal validation set:', np.average(normal_mse))\n",
    "print('Standard deviation of MSE for normal validation set:', np.std(normal_mse))\n",
    "print('Recommended threshold (3x std dev + avg):', (3*np.std(normal_mse)) + np.average(normal_mse))\n",
    "fig, ax = plt.subplots(1,1)\n",
    "ax.hist(normal_mse, bins=20, label='normal', color='blue', alpha=0.7)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracted features from anomaly set. Shape: (59, 3)\n"
     ]
    }
   ],
   "source": [
    "# Extract features from anomaly test set (truncate to length of X test set)\n",
    "anomaly_ops_trunc = anomaly_op_filenames[0:len(normal_mse)]\n",
    "anomaly_features = create_feature_set(anomaly_ops_trunc)\n",
    "print('Extracted features from anomaly set. Shape:', anomaly_features.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average MSE for for anomaly test set: 0.0007622960418707556\n"
     ]
    }
   ],
   "source": [
    "# Calculate MSE from anomaly set\n",
    "predictions = model.predict(anomaly_features)\n",
    "anomaly_mse = np.mean(np.power(anomaly_features - predictions, 2), axis=1)\n",
    "print('Average MSE for for anomaly test set:', np.average(anomaly_mse))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([26.,  0.,  5.,  3.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  1.,  2.,\n",
       "         6.,  2.,  6.,  4.,  0.,  3.,  1.]),\n",
       " array([9.54859542e-06, 1.17009652e-04, 2.24470708e-04, 3.31931765e-04,\n",
       "        4.39392821e-04, 5.46853877e-04, 6.54314934e-04, 7.61775990e-04,\n",
       "        8.69237046e-04, 9.76698103e-04, 1.08415916e-03, 1.19162022e-03,\n",
       "        1.29908127e-03, 1.40654233e-03, 1.51400338e-03, 1.62146444e-03,\n",
       "        1.72892550e-03, 1.83638655e-03, 1.94384761e-03, 2.05130867e-03,\n",
       "        2.15876972e-03]),\n",
       " <a list of 20 Patch objects>)"
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "# Plot histograms of normal validation vs. anomaly sets (MSEs)\n",
    "fig, ax = plt.subplots(1,1)\n",
    "plt.xscale(\"log\")\n",
    "ax.hist(normal_mse, bins=20, label='normal', color='blue', alpha=0.7)\n",
    "ax.hist(anomaly_mse, bins=20, label='anomaly', color='red', alpha=0.7)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average MSE for normal test set: 1.7584283100146983e-05\n"
     ]
    }
   ],
   "source": [
    "# Look at separation using test set\n",
    "predictions = model.predict(x_test)\n",
    "normal_mse = np.mean(np.power(x_test - predictions, 2), axis=1)\n",
    "print('Average MSE for normal test set:', np.average(normal_mse))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([26.,  0.,  5.,  3.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  1.,  2.,\n",
       "         6.,  2.,  6.,  4.,  0.,  3.,  1.]),\n",
       " array([9.54859542e-06, 1.17009652e-04, 2.24470708e-04, 3.31931765e-04,\n",
       "        4.39392821e-04, 5.46853877e-04, 6.54314934e-04, 7.61775990e-04,\n",
       "        8.69237046e-04, 9.76698103e-04, 1.08415916e-03, 1.19162022e-03,\n",
       "        1.29908127e-03, 1.40654233e-03, 1.51400338e-03, 1.62146444e-03,\n",
       "        1.72892550e-03, 1.83638655e-03, 1.94384761e-03, 2.05130867e-03,\n",
       "        2.15876972e-03]),\n",
       " <a list of 20 Patch objects>)"
      ]
     },
     "execution_count": 73,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "# Plot histograms of normal test vs. anomaly sets (MSEs)\n",
    "fig, ax = plt.subplots(1,1)\n",
    "plt.xscale(\"log\")\n",
    "ax.hist(normal_mse, bins=20, label='normal', color='blue', alpha=0.7)\n",
    "ax.hist(anomaly_mse, bins=20, label='anomaly', color='red', alpha=0.7)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "# If we're happy with the performance, save the model\n",
    "model.save(keras_model_name + '.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save a normal and anomaly sample for trying out on the MCU\n",
    "normal_sample = np.genfromtxt(filenames_test[0], delimiter=',')\n",
    "anomaly_sample = np.genfromtxt(anomaly_op_filenames[0], delimiter=',')\n",
    "np.savez(sample_file_name + '.npz', normal_sample=normal_sample, anomaly_sample=anomaly_sample)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save the test dataset for use as a representative dataset\n",
    "np.savez(rep_dataset_name + '.npz', x_test=x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create a classifier (0 = normal, 1 = anomaly)\n",
    "def detect_anomaly(x, model, threshold=0):\n",
    "    input_tensor = x_test[0].reshape(1, -1)\n",
    "    pred = model.predict(input_tensor)\n",
    "    mse = np.mean(np.power(x - pred, 2), axis=1)\n",
    "    if mse > threshold:\n",
    "        return 1\n",
    "    else:\n",
    "        return 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Choose a threshold\n",
    "anomaly_threshold = 3e-05"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0]\n"
     ]
    }
   ],
   "source": [
    "# Perform classification on test set\n",
    "pred_test = [detect_anomaly(x, model, anomaly_threshold) for x in x_test]\n",
    "print(pred_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 1]\n"
     ]
    }
   ],
   "source": [
    "# Perform classification on anomaly set\n",
    "pred_anomaly = [detect_anomaly(x, model, anomaly_threshold) for x in anomaly_features]\n",
    "print(pred_anomaly)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Combine predictions into one long list and create a label list\n",
    "pred = np.array(pred_test + pred_anomaly)\n",
    "labels = ([0] * len(pred_test)) + ([1] * len(pred_anomaly))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[50  9]\n",
      " [12 47]]\n"
     ]
    }
   ],
   "source": [
    "# Create confusion matrix\n",
    "cm = confusion_matrix(labels, pred)\n",
    "print(cm)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(33.0, 0.5, 'Actual')"
      ]
     },
     "execution_count": 89,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Make confusion matrix pretty\n",
    "df_cm = pd.DataFrame(cm, index=['normal', 'anomaly'], columns=['normal', 'anomaly'])\n",
    "plt.figure()\n",
    "sn.heatmap(df_cm, annot=True)\n",
    "plt.title('Confusion Matrix')\n",
    "plt.xlabel('Prediction')\n",
    "plt.ylabel('Actual')"
   ]
  },
  {
   "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}