models/failure_prediction/jnotebooks/stacked_LSTM_Correlation.ipynb

   1 {
   2  "cells": [
   3   {
   4    "cell_type": "markdown",
   5    "metadata": {
   6     "id": "_EvHebagXgW4"
   7    },
   8    "source": [
   9     "Contributors: **Rohit Singh Rathaur, Girish L.** \n",
  10     "\n",
  11     "Copyright [2021](2021) [*Rohit Singh Rathaur, BIT Mesra and Girish L., CIT GUBBI, Karnataka*]\n",
  12     "\n",
  13     "Licensed under the Apache License, Version 2.0 (the \"License\");\n",
  14     "you may not use this file except in compliance with the License.\n",
  15     "You may obtain a copy of the License at\n",
  16     "\n",
  17     "    http://www.apache.org/licenses/LICENSE-2.0\n",
  18     "\n",
  19     "Unless required by applicable law or agreed to in writing, software\n",
  20     "distributed under the License is distributed on an \"AS IS\" BASIS,\n",
  21     "WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n",
  22     "See the License for the specific language governing permissions and\n",
  23     "limitations under the License."
  24    ]
  25   },
  26   {
  27    "cell_type": "code",
  28    "execution_count": null,
  29    "metadata": {
  30     "colab": {
  31      "base_uri": "https://localhost:8080/"
  32     },
  33     "id": "YQ6lT1e2hrx4",
  34     "outputId": "d11aafe6-20e9-4719-a751-0da58234f8c2"
  35    },
  36    "outputs": [],
  37    "source": [
  38     "from google.colab import drive\n",
  39     "drive.mount('/gdrive')"
  40    ]
  41   },
  42   {
  43    "cell_type": "code",
  44    "execution_count": null,
  45    "metadata": {
  46     "id": "tLhroy5BnMnC"
  47    },
  48    "outputs": [],
  49    "source": [
  50     "# Importing libraries\n",
  51     "import tensorflow as tf\n",
  52     "import matplotlib.pyplot as plt\n",
  53     "import matplotlib as mpl\n",
  54     "import pandas as pd\n",
  55     "import numpy as np\n",
  56     "import os"
  57    ]
  58   },
  59   {
  60    "cell_type": "code",
  61    "execution_count": null,
  62    "metadata": {
  63     "colab": {
  64      "base_uri": "https://localhost:8080/",
  65      "height": 419
  66     },
  67     "id": "2-UpMVsSnfCI",
  68     "outputId": "8c28f578-4405-4a27-81e4-c0fd00475b96"
  69    },
  70    "outputs": [],
  71    "source": [
  72     "df_Ellis  = pd.read_csv(\"/gdrive/MyDrive/LFN Anuket/Analysis/data/Final/Ellis_FinalTwoConditionwithOR.csv\")\n",
  73     "df_Ellis"
  74    ]
  75   },
  76   {
  77    "cell_type": "code",
  78    "execution_count": null,
  79    "metadata": {
  80     "colab": {
  81      "base_uri": "https://localhost:8080/",
  82      "height": 293
  83     },
  84     "id": "92xBt43BnjAo",
  85     "outputId": "179123d5-8aec-4dda-a400-9096dec19dd8"
  86    },
  87    "outputs": [],
  88    "source": [
  89     "df_Ellis.plot()"
  90    ]
  91   },
  92   {
  93    "cell_type": "code",
  94    "execution_count": null,
  95    "metadata": {
  96     "colab": {
  97      "base_uri": "https://localhost:8080/",
  98      "height": 879
  99     },
 100     "id": "RSo-aa-SIoBR",
 101     "outputId": "75071f5e-5942-4939-a33b-2dc8780e260c"
 102    },
 103    "outputs": [],
 104    "source": [
 105     "# we show here the hist\n",
 106     "df_Ellis.hist(bins=100,figsize=(20,15))\n",
 107     "#save_fig(\"attribute_histogram_plots\")\n",
 108     "plt.show()"
 109    ]
 110   },
 111   {
 112    "cell_type": "code",
 113    "execution_count": null,
 114    "metadata": {
 115     "colab": {
 116      "base_uri": "https://localhost:8080/",
 117      "height": 634
 118     },
 119     "id": "gggaMJ_2LtFs",
 120     "outputId": "af5ce749-d888-4cae-ddcd-9a5a5ce6e558"
 121    },
 122    "outputs": [],
 123    "source": [
 124     "cpu_system_perc = df_Ellis[['ellis-cpu.system_perc']] \n",
 125     "cpu_system_perc.rolling(12).mean().plot(figsize=(20,10), linewidth=5, fontsize=20) \n",
 126     "plt.xlabel('Timestamp', fontsize=30);"
 127    ]
 128   },
 129   {
 130    "cell_type": "code",
 131    "execution_count": null,
 132    "metadata": {
 133     "colab": {
 134      "base_uri": "https://localhost:8080/",
 135      "height": 634
 136     },
 137     "id": "R_ctvXcQL1Xf",
 138     "outputId": "8beab52c-ffb7-415c-c84c-04ed39cd70cf"
 139    },
 140    "outputs": [],
 141    "source": [
 142     "load_avg_1_min = df_Ellis[['ellis-load.avg_1_min']] \n",
 143     "load_avg_1_min.rolling(12).mean().plot(figsize=(20,10), linewidth=5, fontsize=20) \n",
 144     "plt.xlabel('Timestamp', fontsize=30);"
 145    ]
 146   },
 147   {
 148    "cell_type": "code",
 149    "execution_count": null,
 150    "metadata": {
 151     "colab": {
 152      "base_uri": "https://localhost:8080/",
 153      "height": 634
 154     },
 155     "id": "Gkd5ecCmL6Bw",
 156     "outputId": "a6c042dc-999a-456c-b233-a1e05612e15f"
 157    },
 158    "outputs": [],
 159    "source": [
 160     "cpu_wait_perc = df_Ellis[['ellis-cpu.wait_perc']] \n",
 161     "cpu_wait_perc.rolling(12).mean().plot(figsize=(20,10), linewidth=5, fontsize=20) \n",
 162     "plt.xlabel('Year', fontsize=30);"
 163    ]
 164   },
 165   {
 166    "cell_type": "code",
 167    "execution_count": null,
 168    "metadata": {
 169     "colab": {
 170      "base_uri": "https://localhost:8080/",
 171      "height": 624
 172     },
 173     "id": "EycZrQU0MBSX",
 174     "outputId": "6f325187-ee1b-4122-97f7-80ffdbf149d2"
 175    },
 176    "outputs": [],
 177    "source": [
 178     "df_dg = pd.concat([cpu_system_perc.rolling(12).mean(), load_avg_1_min.rolling(12).mean(),cpu_wait_perc.rolling(12).mean()], axis=1) \n",
 179     "df_dg.plot(figsize=(20,10), linewidth=5, fontsize=20) \n",
 180     "plt.xlabel('Year', fontsize=20); "
 181    ]
 182   },
 183   {
 184    "cell_type": "code",
 185    "execution_count": null,
 186    "metadata": {
 187     "id": "YoQA_MIBMknS"
 188    },
 189    "outputs": [],
 190    "source": []
 191   },
 192   {
 193    "cell_type": "code",
 194    "execution_count": null,
 195    "metadata": {
 196     "colab": {
 197      "base_uri": "https://localhost:8080/",
 198      "height": 710
 199     },
 200     "id": "Pi8UMMitMa3Q",
 201     "outputId": "a44c4d8b-e9d2-4978-fb9e-ae179758e6bc"
 202    },
 203    "outputs": [],
 204    "source": [
 205     "# we establish the corrmartrice\n",
 206     "import seaborn as sns\n",
 207     "color = sns.color_palette()\n",
 208     "sns.set_style('darkgrid')\n",
 209     "\n",
 210     "correaltionMatrice = df_Ellis.corr()\n",
 211     "f, ax = plt.subplots(figsize=(20, 10))\n",
 212     "sns.heatmap(correaltionMatrice, cbar=True, vmin=0, vmax=1, square=True, annot=True);\n",
 213     "plt.show()"
 214    ]
 215   },
 216   {
 217    "cell_type": "code",
 218    "execution_count": null,
 219    "metadata": {
 220     "colab": {
 221      "base_uri": "https://localhost:8080/"
 222     },
 223     "id": "rkYwyKtXMvpy",
 224     "outputId": "c3e3d463-1089-48f7-da23-01a7b92423a5"
 225    },
 226    "outputs": [],
 227    "source": [
 228     "df_Ellis.corrwith(df_Ellis['ellis-load.avg_1_min'])"
 229    ]
 230   },
 231   {
 232    "cell_type": "code",
 233    "execution_count": null,
 234    "metadata": {
 235     "colab": {
 236      "base_uri": "https://localhost:8080/",
 237      "height": 235
 238     },
 239     "id": "5oQK-ddinvCM",
 240     "outputId": "7e9b0a07-d2de-4e42-d45f-572a8abeae9e"
 241    },
 242    "outputs": [],
 243    "source": [
 244     "## ## using multivariate feature \n",
 245     "\n",
 246     "features_3 = ['ellis-cpu.wait_perc', 'ellis-load.avg_1_min', 'ellis-net.in_bytes_sec', 'Label']\n",
 247     "\n",
 248     "features = df_Ellis[features_3]\n",
 249     "features.index = df_Ellis['Timestamp']\n",
 250     "features.head()"
 251    ]
 252   },
 253   {
 254    "cell_type": "code",
 255    "execution_count": null,
 256    "metadata": {
 257     "colab": {
 258      "base_uri": "https://localhost:8080/",
 259      "height": 386
 260     },
 261     "id": "qbqn755fo81g",
 262     "outputId": "f8c5d0a8-c26a-4aa5-a530-957480984106"
 263    },
 264    "outputs": [],
 265    "source": [
 266     "features.plot(subplots=True)"
 267    ]
 268   },
 269   {
 270    "cell_type": "code",
 271    "execution_count": null,
 272    "metadata": {
 273     "id": "jJQD1x9psWCH"
 274    },
 275    "outputs": [],
 276    "source": [
 277     "features = features.values"
 278    ]
 279   },
 280   {
 281    "cell_type": "code",
 282    "execution_count": null,
 283    "metadata": {
 284     "colab": {
 285      "base_uri": "https://localhost:8080/"
 286     },
 287     "id": "xf8WCiykpUzN",
 288     "outputId": "4dca8628-c567-41b0-f262-bc6a3dddb5a7"
 289    },
 290    "outputs": [],
 291    "source": [
 292     "### standardize data\n",
 293     "train_split = 141600\n",
 294     "tf.random.set_seed(13)\n",
 295     "\n",
 296     "### standardize data\n",
 297     "features_mean = features[:train_split].mean()\n",
 298     "features_std = features[:train_split].std()\n",
 299     "features  = (features - features_mean)/ features_std\n",
 300     "\n",
 301     "print(type(features))\n",
 302     "print(features.shape)\n"
 303    ]
 304   },
 305   {
 306    "cell_type": "code",
 307    "execution_count": null,
 308    "metadata": {
 309     "id": "1a0hNDmppnLB"
 310    },
 311    "outputs": [],
 312    "source": [
 313     "### create mutlivariate data\n",
 314     "\n",
 315     "def mutlivariate_data(features , target , start_idx , end_idx , history_size , target_size,\n",
 316     "                      step ,  single_step = False):\n",
 317     "  data = []\n",
 318     "  labels = []\n",
 319     "  start_idx = start_idx + history_size\n",
 320     "  if end_idx is None:\n",
 321     "    end_idx = len(features)- target_size\n",
 322     "  for i in range(start_idx , end_idx ):\n",
 323     "    idxs = range(i-history_size, i, step) ### using step\n",
 324     "    data.append(features[idxs])\n",
 325     "    if single_step:\n",
 326     "      labels.append(target[i+target_size])\n",
 327     "    else:\n",
 328     "      labels.append(target[i:i+target_size])\n",
 329     "\n",
 330     "  return np.array(data) , np.array(labels)"
 331    ]
 332   },
 333   {
 334    "cell_type": "code",
 335    "execution_count": null,
 336    "metadata": {
 337     "colab": {
 338      "base_uri": "https://localhost:8080/"
 339     },
 340     "id": "Z0CivgkitfgE",
 341     "outputId": "7747f2fc-9d3d-4e46-9148-39708b9ae947"
 342    },
 343    "outputs": [],
 344    "source": [
 345     "### generate multivariate data\n",
 346     "\n",
 347     "history = 720\n",
 348     "future_target = 72\n",
 349     "STEP = 6\n",
 350     "\n",
 351     "x_train_ss , y_train_ss = mutlivariate_data(features , features[:, 1], 0, train_split, history,\n",
 352     "                                            future_target, STEP , single_step = True)\n",
 353     "\n",
 354     "x_val_ss , y_val_ss = mutlivariate_data(features , features[:,1] , train_split , None , history ,\n",
 355     "                                        future_target, STEP, single_step = True)\n",
 356     "\n",
 357     "print(x_train_ss.shape , y_train_ss.shape)\n",
 358     "print(x_val_ss.shape , y_val_ss.shape)"
 359    ]
 360   },
 361   {
 362    "cell_type": "code",
 363    "execution_count": null,
 364    "metadata": {
 365     "colab": {
 366      "base_uri": "https://localhost:8080/"
 367     },
 368     "id": "VBdr2epGu3aq",
 369     "outputId": "f3dccf5d-76c2-4c9a-898b-9a221f5a491c"
 370    },
 371    "outputs": [],
 372    "source": [
 373     "## tensorflow dataset\n",
 374     "batch_size = 256\n",
 375     "buffer_size = 10000\n",
 376     "\n",
 377     "train_ss = tf.data.Dataset.from_tensor_slices((x_train_ss, y_train_ss))\n",
 378     "train_ss = train_ss.cache().shuffle(buffer_size).batch(batch_size).repeat()\n",
 379     "\n",
 380     "val_ss = tf.data.Dataset.from_tensor_slices((x_val_ss, y_val_ss))\n",
 381     "val_ss = val_ss.cache().shuffle(buffer_size).batch(batch_size).repeat()\n",
 382     "\n",
 383     "print(train_ss)\n",
 384     "print(val_ss)"
 385    ]
 386   },
 387   {
 388    "cell_type": "code",
 389    "execution_count": null,
 390    "metadata": {
 391     "id": "9eQpwUyGglu_"
 392    },
 393    "outputs": [],
 394    "source": [
 395     "def root_mean_squared_error(y_true, y_pred):\n",
 396     "        return K.sqrt(K.mean(K.square(y_pred - y_true))) "
 397    ]
 398   },
 399   {
 400    "cell_type": "markdown",
 401    "metadata": {
 402     "id": "XkGTfQ46NAlo"
 403    },
 404    "source": [
 405     "## Why Increase Depth?\n",
 406     "Stacking LSTM hidden layers makes the model deeper, more accurately earning the description as a deep learning technique. It is the depth of neural networks that is generally attributed to the success of the approach on a wide range of challenging prediction problems.\n",
 407     "\n",
 408     "As Stacked LSTMs are now a stable technique for challenging sequence prediction problems. A Stacked LSTM architecture is defined as an LSTM model comprised of multiple LSTM layers. An LSTM layer above provides a sequence output rather than a single value output to the LSTM layer below. Specifically, one output per input time step, rather than one output time step for all input time steps.\n",
 409     "\n",
 410     "We created Stacked LSTM model using Keras which is a Python deep learning library. "
 411    ]
 412   },
 413   {
 414    "cell_type": "code",
 415    "execution_count": null,
 416    "metadata": {
 417     "colab": {
 418      "base_uri": "https://localhost:8080/"
 419     },
 420     "id": "1cKtTAzqyiyL",
 421     "outputId": "eed7670c-b3b2-4a20-82bb-7ff2b5f4a0eb"
 422    },
 423    "outputs": [],
 424    "source": [
 425     "from keras.layers import Activation, Dense, Dropout\n",
 426     "### Modelling using LSTM\n",
 427     "steps = 50\n",
 428     "\n",
 429     "EPOCHS =20\n",
 430     "\n",
 431     "single_step_model = tf.keras.models.Sequential()\n",
 432     "\n",
 433     "single_step_model.add(tf.keras.layers.LSTM(32, return_sequences=True, input_shape = x_train_ss.shape[-2:]))\n",
 434     "single_step_model.add(tf.keras.layers.Dropout(0.3))\n",
 435     "single_step_model.add(tf.keras.layers.LSTM(units=100,return_sequences=False))\n",
 436     "single_step_model.add(tf.keras.layers.Dropout(0.2))\n",
 437     "#model.add(Dense(units=1, activation='relu'))\n",
 438     "single_step_model.add(tf.keras.layers.Activation(\"relu\"))\n",
 439     "single_step_model.add(tf.keras.layers.Dense(1))\n",
 440     "single_step_model.compile(optimizer = tf.keras.optimizers.Adam(), loss = 'mae',metrics=[tf.keras.metrics.RootMeanSquaredError(name='rmse')])\n",
 441     "#single_step_model.compile(loss='mse', optimizer='rmsprop')\n",
 442     "single_step_model_history = single_step_model.fit(train_ss, epochs = EPOCHS , \n",
 443     "                                                  steps_per_epoch =steps, validation_data = val_ss,\n",
 444     "                                                  validation_steps = 50)\n",
 445     "\n",
 446     "\n",
 447     "\n",
 448     "single_step_model.summary()\n",
 449     "\n"
 450    ]
 451   },
 452   {
 453    "cell_type": "code",
 454    "execution_count": null,
 455    "metadata": {
 456     "colab": {
 457      "base_uri": "https://localhost:8080/",
 458      "height": 281
 459     },
 460     "id": "Pgev0dgzzBVx",
 461     "outputId": "25c66158-969a-47c4-d35b-9c34abba211e"
 462    },
 463    "outputs": [],
 464    "source": [
 465     "## plot train test loss \n",
 466     "\n",
 467     "def plot_loss(history , title):\n",
 468     "  loss = history.history['loss']\n",
 469     "  val_loss = history.history['val_loss']\n",
 470     "\n",
 471     "  epochs = range(len(loss))\n",
 472     "  plt.figure()\n",
 473     "  plt.plot(epochs, loss , 'b' , label = 'Train Loss')\n",
 474     "  plt.plot(epochs, val_loss , 'r' , label = 'Validation Loss')\n",
 475     "  plt.title(title)\n",
 476     "  plt.legend()\n",
 477     "  plt.grid()\n",
 478     "  plt.show()\n",
 479     "\n",
 480     "plot_loss(single_step_model_history , 'Single Step Training and validation loss')"
 481    ]
 482   },
 483   {
 484    "cell_type": "code",
 485    "execution_count": null,
 486    "metadata": {
 487     "colab": {
 488      "base_uri": "https://localhost:8080/",
 489      "height": 281
 490     },
 491     "id": "EnYf6j4okEoC",
 492     "outputId": "0a474737-67b6-4ff1-d6b2-37b7623a44ce"
 493    },
 494    "outputs": [],
 495    "source": [
 496     "## plot train test loss \n",
 497     "\n",
 498     "def plot_loss(history , title):\n",
 499     "  loss = history.history['rmse']\n",
 500     "  val_loss = history.history['val_rmse']\n",
 501     "\n",
 502     "  epochs = range(len(loss))\n",
 503     "  plt.figure()\n",
 504     "  plt.plot(epochs, loss , 'b' , label = 'Train RMSE')\n",
 505     "  plt.plot(epochs, val_loss , 'r' , label = 'Validation RMSE')\n",
 506     "  plt.title(title)\n",
 507     "  plt.legend()\n",
 508     "  plt.grid()\n",
 509     "  plt.show()\n",
 510     "\n",
 511     "plot_loss(single_step_model_history , 'Single Step Training and validation loss')"
 512    ]
 513   },
 514   {
 515    "cell_type": "code",
 516    "execution_count": null,
 517    "metadata": {
 518     "id": "WMegV8mNAwe_"
 519    },
 520    "outputs": [],
 521    "source": [
 522     "### fucntion to create time steps\n",
 523     "def create_time_steps(length):\n",
 524     "  return list(range(-length,0))\n",
 525     "\n",
 526     "### function to plot time series data\n",
 527     "\n",
 528     "def plot_time_series(plot_data, delta , title):\n",
 529     "  labels = [\"History\" , 'True Future' , 'Model Predcited']\n",
 530     "  marker = ['.-' , 'rx' , 'go']\n",
 531     "  time_steps = create_time_steps(plot_data[0].shape[0])\n",
 532     "\n",
 533     "  if delta:\n",
 534     "    future = delta\n",
 535     "  else:\n",
 536     "    future = 0\n",
 537     "  plt.title(title)\n",
 538     "  for i , x in enumerate(plot_data):\n",
 539     "    if i :\n",
 540     "      plt.plot(future , plot_data[i] , marker[i], markersize = 10 , label = labels[i])\n",
 541     "    else:\n",
 542     "      plt.plot(time_steps, plot_data[i].flatten(), marker[i], label = labels[i])\n",
 543     "  plt.legend()\n",
 544     "  plt.xlim([time_steps[0], (future+5) *2])\n",
 545     "\n",
 546     "  plt.xlabel('Time_Step')\n",
 547     "  return plt"
 548    ]
 549   },
 550   {
 551    "cell_type": "code",
 552    "execution_count": null,
 553    "metadata": {
 554     "id": "q99i2c-9XKF3"
 555    },
 556    "outputs": [],
 557    "source": [
 558     "### Moving window average\n",
 559     "\n",
 560     "def MWA(history):\n",
 561     "  return np.mean(history)"
 562    ]
 563   },
 564   {
 565    "cell_type": "code",
 566    "execution_count": null,
 567    "metadata": {
 568     "colab": {
 569      "base_uri": "https://localhost:8080/",
 570      "height": 1000
 571     },
 572     "id": "xFJT1rZDAUVL",
 573     "outputId": "8e38f8da-d7ac-40ce-eec9-eee2320fd828"
 574    },
 575    "outputs": [],
 576    "source": [
 577     "# plot time series and predicted values\n",
 578     "\n",
 579     "for x, y in val_ss.take(5):\n",
 580     "  plot = plot_time_series([x[0][:, 1].numpy(), y[0].numpy(),\n",
 581     "                    single_step_model.predict(x)[0]], 12,\n",
 582     "                   'Single Step Prediction')\n",
 583     "  plot.show()"
 584    ]
 585   },
 586   {
 587    "cell_type": "markdown",
 588    "metadata": {
 589     "id": "_KXWQVmyCSix"
 590    },
 591    "source": [
 592     "# **MultiStep Forcasting**"
 593    ]
 594   },
 595   {
 596    "cell_type": "code",
 597    "execution_count": null,
 598    "metadata": {
 599     "colab": {
 600      "base_uri": "https://localhost:8080/"
 601     },
 602     "id": "Lu7m2Rr4AbMK",
 603     "outputId": "7b932e10-bb78-4eac-8757-4aa2d915b96d"
 604    },
 605    "outputs": [],
 606    "source": [
 607     "future_target = 72 # 72 future values\n",
 608     "x_train_multi, y_train_multi = mutlivariate_data(features, features[:, 1], 0,\n",
 609     "                                                 train_split, history,\n",
 610     "                                                 future_target, STEP)\n",
 611     "x_val_multi, y_val_multi = mutlivariate_data(features, features[:, 1],\n",
 612     "                                             train_split, None, history,\n",
 613     "                                             future_target, STEP)\n",
 614     "\n",
 615     "print(x_train_multi.shape)\n",
 616     "print(y_train_multi.shape)"
 617    ]
 618   },
 619   {
 620    "cell_type": "code",
 621    "execution_count": null,
 622    "metadata": {
 623     "id": "GLRv5D16CrHj"
 624    },
 625    "outputs": [],
 626    "source": [
 627     "#  TF DATASET\n",
 628     "\n",
 629     "train_data_multi = tf.data.Dataset.from_tensor_slices((x_train_multi, y_train_multi))\n",
 630     "train_data_multi = train_data_multi.cache().shuffle(buffer_size).batch(batch_size).repeat()\n",
 631     "\n",
 632     "val_data_multi = tf.data.Dataset.from_tensor_slices((x_val_multi, y_val_multi))\n",
 633     "val_data_multi = val_data_multi.batch(batch_size).repeat()"
 634    ]
 635   },
 636   {
 637    "cell_type": "code",
 638    "execution_count": null,
 639    "metadata": {
 640     "colab": {
 641      "base_uri": "https://localhost:8080/"
 642     },
 643     "id": "fjXexah9C8yg",
 644     "outputId": "72c81185-3db8-4d45-f1f2-9bf5861c3a3e"
 645    },
 646    "outputs": [],
 647    "source": [
 648     "print(train_data_multi)\n",
 649     "print(val_data_multi)"
 650    ]
 651   },
 652   {
 653    "cell_type": "code",
 654    "execution_count": null,
 655    "metadata": {
 656     "colab": {
 657      "base_uri": "https://localhost:8080/",
 658      "height": 385
 659     },
 660     "id": "7mtLZ6S-DPU-",
 661     "outputId": "a4a392fe-73a5-44f7-ae77-003e7c46a1ad"
 662    },
 663    "outputs": [],
 664    "source": [
 665     "#plotting function\n",
 666     "def multi_step_plot(history, true_future, prediction):\n",
 667     "  plt.figure(figsize=(12, 6))\n",
 668     "  num_in = create_time_steps(len(history))\n",
 669     "  num_out = len(true_future)\n",
 670     "  plt.grid()\n",
 671     "  plt.plot(num_in, np.array(history[:, 1]), label='History')\n",
 672     "  plt.plot(np.arange(num_out)/STEP, np.array(true_future), 'bo',\n",
 673     "           label='True Future')\n",
 674     "  if prediction.any():\n",
 675     "    plt.plot(np.arange(num_out)/STEP, np.array(prediction), 'ro',\n",
 676     "             label='Predicted Future')\n",
 677     "  plt.legend(loc='upper left')\n",
 678     "  plt.show()\n",
 679     "  \n",
 680     "\n",
 681     "\n",
 682     "for x, y in train_data_multi.take(1):\n",
 683     "  multi_step_plot(x[0], y[0], np.array([0]))"
 684    ]
 685   },
 686   {
 687    "cell_type": "code",
 688    "execution_count": null,
 689    "metadata": {
 690     "colab": {
 691      "base_uri": "https://localhost:8080/"
 692     },
 693     "id": "snN_Flr5DWQN",
 694     "outputId": "8b881aff-b91b-4394-bc3d-e52261e546db"
 695    },
 696    "outputs": [],
 697    "source": [
 698     "multi_step_model = tf.keras.models.Sequential()\n",
 699     "multi_step_model.add(tf.keras.layers.LSTM(32,\n",
 700     "                                          return_sequences=True,\n",
 701     "                                          input_shape=x_train_multi.shape[-2:]))\n",
 702     "multi_step_model.add(tf.keras.layers.Dropout(0.2))\n",
 703     "multi_step_model.add(tf.keras.layers.LSTM(units=100,return_sequences=False))\n",
 704     "multi_step_model.add(tf.keras.layers.Dropout(0.2))\n",
 705     "#model.add(Dense(units=1, activation='relu'))\n",
 706     "multi_step_model.add(tf.keras.layers.Activation(\"relu\"))\n",
 707     "#aDD dropout layer (0.3)\n",
 708     "multi_step_model.add(tf.keras.layers.Dense(72)) # for 72 outputs\n",
 709     "\n",
 710     "multi_step_model.compile(optimizer=tf.keras.optimizers.RMSprop(clipvalue=1.0), loss='mae',metrics=[tf.keras.metrics.RootMeanSquaredError(name='rmse')])\n",
 711     "\n",
 712     "multi_step_history = multi_step_model.fit(train_data_multi, epochs=EPOCHS,\n",
 713     "                                          steps_per_epoch=steps,\n",
 714     "                                          validation_data=val_data_multi,\n",
 715     "                                          validation_steps=50)"
 716    ]
 717   },
 718   {
 719    "cell_type": "code",
 720    "execution_count": null,
 721    "metadata": {
 722     "colab": {
 723      "base_uri": "https://localhost:8080/"
 724     },
 725     "id": "Ay5m27doDsTt",
 726     "outputId": "70f65cf2-a2fd-4a1e-b9e0-049430410a1d"
 727    },
 728    "outputs": [],
 729    "source": [
 730     "plot_loss(multi_step_history, 'Multi-Step Training and validation loss')"
 731    ]
 732   },
 733   {
 734    "cell_type": "code",
 735    "execution_count": null,
 736    "metadata": {
 737     "colab": {
 738      "base_uri": "https://localhost:8080/",
 739      "height": 1000
 740     },
 741     "id": "6ZFP49W4D2wp",
 742     "outputId": "f2d25889-8aa1-4405-d607-1ed72a0df675"
 743    },
 744    "outputs": [],
 745    "source": [
 746     "for x, y in val_data_multi.take(5):\n",
 747     "  multi_step_plot(x[0], y[0], multi_step_model.predict(x)[0])"
 748    ]
 749   },
 750   {
 751    "cell_type": "code",
 752    "execution_count": null,
 753    "metadata": {
 754     "colab": {
 755      "base_uri": "https://localhost:8080/"
 756     },
 757     "id": "DNKMjVoAVqZP",
 758     "outputId": "d853fe0e-6ca4-411f-aa83-248f61309efe"
 759    },
 760    "outputs": [],
 761    "source": [
 762     "scores = multi_step_model.evaluate(x_train_multi, y_train_multi, verbose=1, batch_size=200)\n",
 763     "print('MAE: {}'.format(scores[1]))"
 764    ]
 765   },
 766   {
 767    "cell_type": "code",
 768    "execution_count": null,
 769    "metadata": {
 770     "colab": {
 771      "base_uri": "https://localhost:8080/"
 772     },
 773     "id": "YXcsNZ8yu9Ay",
 774     "outputId": "0b7ee485-e356-4c5b-8a9c-257d8803d189"
 775    },
 776    "outputs": [],
 777    "source": [
 778     "scores_test = multi_step_model.evaluate(x_val_multi, y_val_multi, verbose=1, batch_size=200)\n",
 779     "print('MAE: {}'.format(scores[1]))\n"
 780    ]
 781   },
 782   {
 783    "cell_type": "code",
 784    "execution_count": null,
 785    "metadata": {
 786     "colab": {
 787      "base_uri": "https://localhost:8080/"
 788     },
 789     "id": "uCFgbZEOvZ9A",
 790     "outputId": "46302cac-3d92-4970-c100-e5c3a13847f0"
 791    },
 792    "outputs": [],
 793    "source": [
 794     "y_pred_test = multi_step_model.predict(x_val_multi, verbose=0)\n",
 795     "\n",
 796     "plt.figure(figsize=(10,5))\n",
 797     "plt.plot(y_pred_test)\n",
 798     "plt.plot(y_val_multi)\n",
 799     "plt.ylabel(\"RUL\")\n",
 800     "plt.xlabel(\"Unit Number\")\n",
 801     "plt.legend(loc='upper left')\n",
 802     "plt.show()"
 803    ]
 804   }
 805  ],
 806  "metadata": {
 807   "accelerator": "GPU",
 808   "colab": {
 809    "collapsed_sections": [],
 810    "name": "stacked-LSTM_Correlation.ipynb",
 811    "provenance": []
 812   },
 813   "kernelspec": {
 814    "display_name": "Python 3 (ipykernel)",
 815    "language": "python",
 816    "name": "python3"
 817   },
 818   "language_info": {
 819    "codemirror_mode": {
 820     "name": "ipython",
 821     "version": 3
 822    },
 823    "file_extension": ".py",
 824    "mimetype": "text/x-python",
 825    "name": "python",
 826    "nbconvert_exporter": "python",
 827    "pygments_lexer": "ipython3",
 828    "version": "3.9.7"
 829   }
 830  },
 831  "nbformat": 4,
 832  "nbformat_minor": 1
 833 }