emotion detection/Transfer Learning - Neural Networks .ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "np.random.seed(888)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "from IPython.display import SVG\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation, Dropout\n",
    "from keras.utils import model_to_dot\n",
    "from keras.callbacks.callbacks import EarlyStopping"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Read train, valid, test data set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_data = np.load(\"data/train_data.npy\")\n",
    "train_label = np.load(\"data/train_label.npy\")\n",
    "\n",
    "valid_data = np.load(\"data/valid_data.npy\")\n",
    "valid_label = np.load(\"data/valid_label.npy\")\n",
    "\n",
    "test_data = np.load(\"data/test_data.npy\")\n",
    "test_label = np.load(\"data/test_label.npy\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "monitor = EarlyStopping(monitor='val_loss', min_delta=1e-3, patience=5, verbose=1, mode='auto')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Neural Network 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_1 = Sequential()\n",
    "model_1.add(Dense(4, input_shape=(2304,)))\n",
    "model_1.add(Activation('softmax'))\n",
    "model_1.compile(optimizer='adam',\n",
    "                loss='categorical_crossentropy',\n",
    "                metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/svg+xml": [
       "<svg height=\"255pt\" viewBox=\"0.00 0.00 175.24 191.00\" width=\"234pt\" xmlns=\"http://www.w3.org/2000/svg\" xmlns:xlink=\"http://www.w3.org/1999/xlink\">\n",
       "<g class=\"graph\" id=\"graph0\" transform=\"scale(0.75 0.75) rotate(0) translate(4 187)\">\n",
       "<title>G</title>\n",
       "<polygon fill=\"white\" points=\"-4,4 -4,-187 171.24,-187 171.24,4 -4,4\" stroke=\"transparent\"/>\n",
       "<!-- 5193002680 -->\n",
       "<g class=\"node\" id=\"node1\">\n",
       "<title>5193002680</title>\n",
       "<polygon fill=\"none\" points=\"0,-146.5 0,-182.5 167.24,-182.5 167.24,-146.5 0,-146.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.62\" y=\"-160.3\">dense_1_input: InputLayer</text>\n",
       "</g>\n",
       "<!-- 5193000944 -->\n",
       "<g class=\"node\" id=\"node2\">\n",
       "<title>5193000944</title>\n",
       "<polygon fill=\"none\" points=\"31.49,-73.5 31.49,-109.5 135.75,-109.5 135.75,-73.5 31.49,-73.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.62\" y=\"-87.3\">dense_1: Dense</text>\n",
       "</g>\n",
       "<!-- 5193002680&#45;&gt;5193000944 -->\n",
       "<g class=\"edge\" id=\"edge1\">\n",
       "<title>5193002680-&gt;5193000944</title>\n",
       "<path d=\"M83.62,-146.31C83.62,-138.29 83.62,-128.55 83.62,-119.57\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"87.12,-119.53 83.62,-109.53 80.12,-119.53 87.12,-119.53\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 4558215656 -->\n",
       "<g class=\"node\" id=\"node3\">\n",
       "<title>4558215656</title>\n",
       "<polygon fill=\"none\" points=\"8.16,-0.5 8.16,-36.5 159.08,-36.5 159.08,-0.5 8.16,-0.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.62\" y=\"-14.3\">activation_1: Activation</text>\n",
       "</g>\n",
       "<!-- 5193000944&#45;&gt;4558215656 -->\n",
       "<g class=\"edge\" id=\"edge2\">\n",
       "<title>5193000944-&gt;4558215656</title>\n",
       "<path d=\"M83.62,-73.31C83.62,-65.29 83.62,-55.55 83.62,-46.57\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"87.12,-46.53 83.62,-36.53 80.12,-46.53 87.12,-46.53\" stroke=\"black\"/>\n",
       "</g>\n",
       "</g>\n",
       "</svg>"
      ],
      "text/plain": [
       "<IPython.core.display.SVG object>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "SVG(model_to_dot(model_1).create(prog='dot', format='svg'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 105697 samples, validate on 13213 samples\n",
      "Epoch 1/50\n",
      "105697/105697 [==============================] - 2s 14us/step - loss: 1.0453 - accuracy: 0.5541 - val_loss: 0.9755 - val_accuracy: 0.5753\n",
      "Epoch 2/50\n",
      "105697/105697 [==============================] - 2s 15us/step - loss: 0.9706 - accuracy: 0.5783 - val_loss: 0.9612 - val_accuracy: 0.5823\n",
      "Epoch 3/50\n",
      "105697/105697 [==============================] - 1s 13us/step - loss: 0.9602 - accuracy: 0.5839 - val_loss: 0.9561 - val_accuracy: 0.5859\n",
      "Epoch 4/50\n",
      "105697/105697 [==============================] - 2s 15us/step - loss: 0.9550 - accuracy: 0.5863 - val_loss: 0.9539 - val_accuracy: 0.5877\n",
      "Epoch 5/50\n",
      "105697/105697 [==============================] - 2s 15us/step - loss: 0.9514 - accuracy: 0.5889 - val_loss: 0.9528 - val_accuracy: 0.5883\n",
      "Epoch 6/50\n",
      "105697/105697 [==============================] - 2s 15us/step - loss: 0.9487 - accuracy: 0.5902 - val_loss: 0.9524 - val_accuracy: 0.5890\n",
      "Epoch 7/50\n",
      "105697/105697 [==============================] - 2s 18us/step - loss: 0.9465 - accuracy: 0.5912 - val_loss: 0.9509 - val_accuracy: 0.5887\n",
      "Epoch 8/50\n",
      "105697/105697 [==============================] - 2s 18us/step - loss: 0.9445 - accuracy: 0.5929 - val_loss: 0.9506 - val_accuracy: 0.5881\n",
      "Epoch 9/50\n",
      "105697/105697 [==============================] - 2s 15us/step - loss: 0.9431 - accuracy: 0.5922 - val_loss: 0.9512 - val_accuracy: 0.5862\n",
      "Epoch 10/50\n",
      "105697/105697 [==============================] - 1s 14us/step - loss: 0.9416 - accuracy: 0.5934 - val_loss: 0.9504 - val_accuracy: 0.5883\n",
      "Epoch 11/50\n",
      "105697/105697 [==============================] - 1s 14us/step - loss: 0.9405 - accuracy: 0.5938 - val_loss: 0.9500 - val_accuracy: 0.5878\n",
      "Epoch 12/50\n",
      "105697/105697 [==============================] - 2s 15us/step - loss: 0.9391 - accuracy: 0.5952 - val_loss: 0.9505 - val_accuracy: 0.5871\n",
      "Epoch 00012: early stopping\n"
     ]
    }
   ],
   "source": [
    "history_1 = model_1.fit(x = train_data, y = train_label, \n",
    "                        validation_data = (valid_data, valid_label),\n",
    "                        callbacks = [monitor],\n",
    "                        epochs = 50, batch_size = 128, \n",
    "                        shuffle = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "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 training & validation accuracy values\n",
    "plt.plot(history_1.history['accuracy'])\n",
    "plt.plot(history_1.history['val_accuracy'])\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "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 training & validation loss values\n",
    "plt.plot(history_1.history['loss'])\n",
    "plt.plot(history_1.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": "markdown",
   "metadata": {},
   "source": [
    "## Neural Network 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_2 = Sequential()\n",
    "model_2.add(Dense(128, input_shape=(2304,), activation='tanh'))\n",
    "model_2.add(Dense(4))\n",
    "model_2.add(Activation('softmax'))\n",
    "model_2.compile(optimizer='adam',\n",
    "              loss='categorical_crossentropy',\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/svg+xml": [
       "<svg height=\"352pt\" viewBox=\"0.00 0.00 175.24 264.00\" width=\"234pt\" xmlns=\"http://www.w3.org/2000/svg\" xmlns:xlink=\"http://www.w3.org/1999/xlink\">\n",
       "<g class=\"graph\" id=\"graph0\" transform=\"scale(0.75 0.75) rotate(0) translate(4 260)\">\n",
       "<title>G</title>\n",
       "<polygon fill=\"white\" points=\"-4,4 -4,-260 171.24,-260 171.24,4 -4,4\" stroke=\"transparent\"/>\n",
       "<!-- 4627319944 -->\n",
       "<g class=\"node\" id=\"node1\">\n",
       "<title>4627319944</title>\n",
       "<polygon fill=\"none\" points=\"0,-219.5 0,-255.5 167.24,-255.5 167.24,-219.5 0,-219.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.62\" y=\"-233.3\">dense_2_input: InputLayer</text>\n",
       "</g>\n",
       "<!-- 6416172648 -->\n",
       "<g class=\"node\" id=\"node2\">\n",
       "<title>6416172648</title>\n",
       "<polygon fill=\"none\" points=\"31.49,-146.5 31.49,-182.5 135.75,-182.5 135.75,-146.5 31.49,-146.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.62\" y=\"-160.3\">dense_2: Dense</text>\n",
       "</g>\n",
       "<!-- 4627319944&#45;&gt;6416172648 -->\n",
       "<g class=\"edge\" id=\"edge1\">\n",
       "<title>4627319944-&gt;6416172648</title>\n",
       "<path d=\"M83.62,-219.31C83.62,-211.29 83.62,-201.55 83.62,-192.57\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"87.12,-192.53 83.62,-182.53 80.12,-192.53 87.12,-192.53\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 4625904640 -->\n",
       "<g class=\"node\" id=\"node3\">\n",
       "<title>4625904640</title>\n",
       "<polygon fill=\"none\" points=\"31.49,-73.5 31.49,-109.5 135.75,-109.5 135.75,-73.5 31.49,-73.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.62\" y=\"-87.3\">dense_3: Dense</text>\n",
       "</g>\n",
       "<!-- 6416172648&#45;&gt;4625904640 -->\n",
       "<g class=\"edge\" id=\"edge2\">\n",
       "<title>6416172648-&gt;4625904640</title>\n",
       "<path d=\"M83.62,-146.31C83.62,-138.29 83.62,-128.55 83.62,-119.57\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"87.12,-119.53 83.62,-109.53 80.12,-119.53 87.12,-119.53\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 4627130576 -->\n",
       "<g class=\"node\" id=\"node4\">\n",
       "<title>4627130576</title>\n",
       "<polygon fill=\"none\" points=\"8.16,-0.5 8.16,-36.5 159.08,-36.5 159.08,-0.5 8.16,-0.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.62\" y=\"-14.3\">activation_2: Activation</text>\n",
       "</g>\n",
       "<!-- 4625904640&#45;&gt;4627130576 -->\n",
       "<g class=\"edge\" id=\"edge3\">\n",
       "<title>4625904640-&gt;4627130576</title>\n",
       "<path d=\"M83.62,-73.31C83.62,-65.29 83.62,-55.55 83.62,-46.57\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"87.12,-46.53 83.62,-36.53 80.12,-46.53 87.12,-46.53\" stroke=\"black\"/>\n",
       "</g>\n",
       "</g>\n",
       "</svg>"
      ],
      "text/plain": [
       "<IPython.core.display.SVG object>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "SVG(model_to_dot(model_2).create(prog='dot', format='svg'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 105697 samples, validate on 13213 samples\n",
      "Epoch 1/50\n",
      "105697/105697 [==============================] - 4s 35us/step - loss: 0.9817 - accuracy: 0.5717 - val_loss: 0.9703 - val_accuracy: 0.5777\n",
      "Epoch 2/50\n",
      "105697/105697 [==============================] - 4s 33us/step - loss: 0.9555 - accuracy: 0.5858 - val_loss: 0.9545 - val_accuracy: 0.5844\n",
      "Epoch 3/50\n",
      "105697/105697 [==============================] - 3s 33us/step - loss: 0.9486 - accuracy: 0.5898 - val_loss: 0.9551 - val_accuracy: 0.5841\n",
      "Epoch 4/50\n",
      "105697/105697 [==============================] - 4s 34us/step - loss: 0.9449 - accuracy: 0.5926 - val_loss: 0.9571 - val_accuracy: 0.5829\n",
      "Epoch 5/50\n",
      "105697/105697 [==============================] - 4s 33us/step - loss: 0.9429 - accuracy: 0.5931 - val_loss: 0.9541 - val_accuracy: 0.5859\n",
      "Epoch 6/50\n",
      "105697/105697 [==============================] - 3s 33us/step - loss: 0.9405 - accuracy: 0.5939 - val_loss: 0.9591 - val_accuracy: 0.5825\n",
      "Epoch 7/50\n",
      "105697/105697 [==============================] - 3s 32us/step - loss: 0.9383 - accuracy: 0.5946 - val_loss: 0.9570 - val_accuracy: 0.5841\n",
      "Epoch 00007: early stopping\n"
     ]
    }
   ],
   "source": [
    "history_2 = model_2.fit(x = train_data, y = train_label, \n",
    "                        validation_data = (valid_data, valid_label),\n",
    "                        callbacks = [monitor],\n",
    "                        epochs = 50, batch_size = 128, \n",
    "                        shuffle = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "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 training & validation accuracy values\n",
    "plt.plot(history_2.history['accuracy'])\n",
    "plt.plot(history_2.history['val_accuracy'])\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "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 training & validation loss values\n",
    "plt.plot(history_2.history['loss'])\n",
    "plt.plot(history_2.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": "markdown",
   "metadata": {},
   "source": [
    "## Neural Network 3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_3 = Sequential()\n",
    "model_3.add(Dense(128, input_shape=(2304,), activation='tanh'))\n",
    "model_3.add(Dropout(0.5))\n",
    "model_3.add(Dense(4))\n",
    "model_3.add(Activation('softmax'))\n",
    "model_3.compile(optimizer='adam',\n",
    "              loss='categorical_crossentropy',\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/svg+xml": [
       "<svg height=\"449pt\" viewBox=\"0.00 0.00 175.24 337.00\" width=\"234pt\" xmlns=\"http://www.w3.org/2000/svg\" xmlns:xlink=\"http://www.w3.org/1999/xlink\">\n",
       "<g class=\"graph\" id=\"graph0\" transform=\"scale(0.75 0.75) rotate(0) translate(4 333)\">\n",
       "<title>G</title>\n",
       "<polygon fill=\"white\" points=\"-4,4 -4,-333 171.24,-333 171.24,4 -4,4\" stroke=\"transparent\"/>\n",
       "<!-- 4627655872 -->\n",
       "<g class=\"node\" id=\"node1\">\n",
       "<title>4627655872</title>\n",
       "<polygon fill=\"none\" points=\"0,-292.5 0,-328.5 167.24,-328.5 167.24,-292.5 0,-292.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.62\" y=\"-306.3\">dense_4_input: InputLayer</text>\n",
       "</g>\n",
       "<!-- 4627597184 -->\n",
       "<g class=\"node\" id=\"node2\">\n",
       "<title>4627597184</title>\n",
       "<polygon fill=\"none\" points=\"31.49,-219.5 31.49,-255.5 135.75,-255.5 135.75,-219.5 31.49,-219.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.62\" y=\"-233.3\">dense_4: Dense</text>\n",
       "</g>\n",
       "<!-- 4627655872&#45;&gt;4627597184 -->\n",
       "<g class=\"edge\" id=\"edge1\">\n",
       "<title>4627655872-&gt;4627597184</title>\n",
       "<path d=\"M83.62,-292.31C83.62,-284.29 83.62,-274.55 83.62,-265.57\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"87.12,-265.53 83.62,-255.53 80.12,-265.53 87.12,-265.53\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 4627598136 -->\n",
       "<g class=\"node\" id=\"node3\">\n",
       "<title>4627598136</title>\n",
       "<polygon fill=\"none\" points=\"19.82,-146.5 19.82,-182.5 147.42,-182.5 147.42,-146.5 19.82,-146.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.62\" y=\"-160.3\">dropout_1: Dropout</text>\n",
       "</g>\n",
       "<!-- 4627597184&#45;&gt;4627598136 -->\n",
       "<g class=\"edge\" id=\"edge2\">\n",
       "<title>4627597184-&gt;4627598136</title>\n",
       "<path d=\"M83.62,-219.31C83.62,-211.29 83.62,-201.55 83.62,-192.57\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"87.12,-192.53 83.62,-182.53 80.12,-192.53 87.12,-192.53\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 4627655480 -->\n",
       "<g class=\"node\" id=\"node4\">\n",
       "<title>4627655480</title>\n",
       "<polygon fill=\"none\" points=\"31.49,-73.5 31.49,-109.5 135.75,-109.5 135.75,-73.5 31.49,-73.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.62\" y=\"-87.3\">dense_5: Dense</text>\n",
       "</g>\n",
       "<!-- 4627598136&#45;&gt;4627655480 -->\n",
       "<g class=\"edge\" id=\"edge3\">\n",
       "<title>4627598136-&gt;4627655480</title>\n",
       "<path d=\"M83.62,-146.31C83.62,-138.29 83.62,-128.55 83.62,-119.57\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"87.12,-119.53 83.62,-109.53 80.12,-119.53 87.12,-119.53\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 4627655312 -->\n",
       "<g class=\"node\" id=\"node5\">\n",
       "<title>4627655312</title>\n",
       "<polygon fill=\"none\" points=\"8.16,-0.5 8.16,-36.5 159.08,-36.5 159.08,-0.5 8.16,-0.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.62\" y=\"-14.3\">activation_3: Activation</text>\n",
       "</g>\n",
       "<!-- 4627655480&#45;&gt;4627655312 -->\n",
       "<g class=\"edge\" id=\"edge4\">\n",
       "<title>4627655480-&gt;4627655312</title>\n",
       "<path d=\"M83.62,-73.31C83.62,-65.29 83.62,-55.55 83.62,-46.57\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"87.12,-46.53 83.62,-36.53 80.12,-46.53 87.12,-46.53\" stroke=\"black\"/>\n",
       "</g>\n",
       "</g>\n",
       "</svg>"
      ],
      "text/plain": [
       "<IPython.core.display.SVG object>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "SVG(model_to_dot(model_3).create(prog='dot', format='svg'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 105697 samples, validate on 13213 samples\n",
      "Epoch 1/50\n",
      "105697/105697 [==============================] - 4s 37us/step - loss: 0.9931 - accuracy: 0.5676 - val_loss: 0.9615 - val_accuracy: 0.5819\n",
      "Epoch 2/50\n",
      "105697/105697 [==============================] - 3s 32us/step - loss: 0.9667 - accuracy: 0.5804 - val_loss: 0.9542 - val_accuracy: 0.5839\n",
      "Epoch 3/50\n",
      "105697/105697 [==============================] - 4s 33us/step - loss: 0.9602 - accuracy: 0.5850 - val_loss: 0.9542 - val_accuracy: 0.5858\n",
      "Epoch 4/50\n",
      "105697/105697 [==============================] - 4s 33us/step - loss: 0.9567 - accuracy: 0.5865 - val_loss: 0.9533 - val_accuracy: 0.5876\n",
      "Epoch 5/50\n",
      "105697/105697 [==============================] - 4s 34us/step - loss: 0.9543 - accuracy: 0.5886 - val_loss: 0.9572 - val_accuracy: 0.5838\n",
      "Epoch 6/50\n",
      "105697/105697 [==============================] - 3s 33us/step - loss: 0.9515 - accuracy: 0.5895 - val_loss: 0.9591 - val_accuracy: 0.5801\n",
      "Epoch 7/50\n",
      "105697/105697 [==============================] - 3s 33us/step - loss: 0.9492 - accuracy: 0.5908 - val_loss: 0.9550 - val_accuracy: 0.5835\n",
      "Epoch 00007: early stopping\n"
     ]
    }
   ],
   "source": [
    "history_3 = model_3.fit(x = train_data, y = train_label, \n",
    "                        validation_data = (valid_data, valid_label),\n",
    "                        callbacks = [monitor],\n",
    "                        epochs = 50, batch_size = 128, \n",
    "                        shuffle = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "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 training & validation accuracy values\n",
    "plt.plot(history_3.history['accuracy'])\n",
    "plt.plot(history_3.history['val_accuracy'])\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "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 training & validation loss values\n",
    "plt.plot(history_3.history['loss'])\n",
    "plt.plot(history_3.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": "markdown",
   "metadata": {},
   "source": [
    "## Model selection\n",
    "\n",
    "Choose model 1 as the final model becasuse it has highest validation accuracy (58.71%)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Error Analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "    <div class=\"bk-root\">\n",
       "        <a href=\"https://bokeh.pydata.org\" target=\"_blank\" class=\"bk-logo bk-logo-small bk-logo-notebook\"></a>\n",
       "        <span id=\"1001\">Loading BokehJS ...</span>\n",
       "    </div>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "(function(root) {\n",
       "  function now() {\n",
       "    return new Date();\n",
       "  }\n",
       "\n",
       "  var force = true;\n",
       "\n",
       "  if (typeof root._bokeh_onload_callbacks === \"undefined\" || force === true) {\n",
       "    root._bokeh_onload_callbacks = [];\n",
       "    root._bokeh_is_loading = undefined;\n",
       "  }\n",
       "\n",
       "  var JS_MIME_TYPE = 'application/javascript';\n",
       "  var HTML_MIME_TYPE = 'text/html';\n",
       "  var EXEC_MIME_TYPE = 'application/vnd.bokehjs_exec.v0+json';\n",
       "  var CLASS_NAME = 'output_bokeh rendered_html';\n",
       "\n",
       "  /**\n",
       "   * Render data to the DOM node\n",
       "   */\n",
       "  function render(props, node) {\n",
       "    var script = document.createElement(\"script\");\n",
       "    node.appendChild(script);\n",
       "  }\n",
       "\n",
       "  /**\n",
       "   * Handle when an output is cleared or removed\n",
       "   */\n",
       "  function handleClearOutput(event, handle) {\n",
       "    var cell = handle.cell;\n",
       "\n",
       "    var id = cell.output_area._bokeh_element_id;\n",
       "    var server_id = cell.output_area._bokeh_server_id;\n",
       "    // Clean up Bokeh references\n",
       "    if (id != null && id in Bokeh.index) {\n",
       "      Bokeh.index[id].model.document.clear();\n",
       "      delete Bokeh.index[id];\n",
       "    }\n",
       "\n",
       "    if (server_id !== undefined) {\n",
       "      // Clean up Bokeh references\n",
       "      var cmd = \"from bokeh.io.state import curstate; print(curstate().uuid_to_server['\" + server_id + \"'].get_sessions()[0].document.roots[0]._id)\";\n",
       "      cell.notebook.kernel.execute(cmd, {\n",
       "        iopub: {\n",
       "          output: function(msg) {\n",
       "            var id = msg.content.text.trim();\n",
       "            if (id in Bokeh.index) {\n",
       "              Bokeh.index[id].model.document.clear();\n",
       "              delete Bokeh.index[id];\n",
       "            }\n",
       "          }\n",
       "        }\n",
       "      });\n",
       "      // Destroy server and session\n",
       "      var cmd = \"import bokeh.io.notebook as ion; ion.destroy_server('\" + server_id + \"')\";\n",
       "      cell.notebook.kernel.execute(cmd);\n",
       "    }\n",
       "  }\n",
       "\n",
       "  /**\n",
       "   * Handle when a new output is added\n",
       "   */\n",
       "  function handleAddOutput(event, handle) {\n",
       "    var output_area = handle.output_area;\n",
       "    var output = handle.output;\n",
       "\n",
       "    // limit handleAddOutput to display_data with EXEC_MIME_TYPE content only\n",
       "    if ((output.output_type != \"display_data\") || (!output.data.hasOwnProperty(EXEC_MIME_TYPE))) {\n",
       "      return\n",
       "    }\n",
       "\n",
       "    var toinsert = output_area.element.find(\".\" + CLASS_NAME.split(' ')[0]);\n",
       "\n",
       "    if (output.metadata[EXEC_MIME_TYPE][\"id\"] !== undefined) {\n",
       "      toinsert[toinsert.length - 1].firstChild.textContent = output.data[JS_MIME_TYPE];\n",
       "      // store reference to embed id on output_area\n",
       "      output_area._bokeh_element_id = output.metadata[EXEC_MIME_TYPE][\"id\"];\n",
       "    }\n",
       "    if (output.metadata[EXEC_MIME_TYPE][\"server_id\"] !== undefined) {\n",
       "      var bk_div = document.createElement(\"div\");\n",
       "      bk_div.innerHTML = output.data[HTML_MIME_TYPE];\n",
       "      var script_attrs = bk_div.children[0].attributes;\n",
       "      for (var i = 0; i < script_attrs.length; i++) {\n",
       "        toinsert[toinsert.length - 1].firstChild.setAttribute(script_attrs[i].name, script_attrs[i].value);\n",
       "      }\n",
       "      // store reference to server id on output_area\n",
       "      output_area._bokeh_server_id = output.metadata[EXEC_MIME_TYPE][\"server_id\"];\n",
       "    }\n",
       "  }\n",
       "\n",
       "  function register_renderer(events, OutputArea) {\n",
       "\n",
       "    function append_mime(data, metadata, element) {\n",
       "      // create a DOM node to render to\n",
       "      var toinsert = this.create_output_subarea(\n",
       "        metadata,\n",
       "        CLASS_NAME,\n",
       "        EXEC_MIME_TYPE\n",
       "      );\n",
       "      this.keyboard_manager.register_events(toinsert);\n",
       "      // Render to node\n",
       "      var props = {data: data, metadata: metadata[EXEC_MIME_TYPE]};\n",
       "      render(props, toinsert[toinsert.length - 1]);\n",
       "      element.append(toinsert);\n",
       "      return toinsert\n",
       "    }\n",
       "\n",
       "    /* Handle when an output is cleared or removed */\n",
       "    events.on('clear_output.CodeCell', handleClearOutput);\n",
       "    events.on('delete.Cell', handleClearOutput);\n",
       "\n",
       "    /* Handle when a new output is added */\n",
       "    events.on('output_added.OutputArea', handleAddOutput);\n",
       "\n",
       "    /**\n",
       "     * Register the mime type and append_mime function with output_area\n",
       "     */\n",
       "    OutputArea.prototype.register_mime_type(EXEC_MIME_TYPE, append_mime, {\n",
       "      /* Is output safe? */\n",
       "      safe: true,\n",
       "      /* Index of renderer in `output_area.display_order` */\n",
       "      index: 0\n",
       "    });\n",
       "  }\n",
       "\n",
       "  // register the mime type if in Jupyter Notebook environment and previously unregistered\n",
       "  if (root.Jupyter !== undefined) {\n",
       "    var events = require('base/js/events');\n",
       "    var OutputArea = require('notebook/js/outputarea').OutputArea;\n",
       "\n",
       "    if (OutputArea.prototype.mime_types().indexOf(EXEC_MIME_TYPE) == -1) {\n",
       "      register_renderer(events, OutputArea);\n",
       "    }\n",
       "  }\n",
       "\n",
       "  \n",
       "  if (typeof (root._bokeh_timeout) === \"undefined\" || force === true) {\n",
       "    root._bokeh_timeout = Date.now() + 5000;\n",
       "    root._bokeh_failed_load = false;\n",
       "  }\n",
       "\n",
       "  var NB_LOAD_WARNING = {'data': {'text/html':\n",
       "     \"<div style='background-color: #fdd'>\\n\"+\n",
       "     \"<p>\\n\"+\n",
       "     \"BokehJS does not appear to have successfully loaded. If loading BokehJS from CDN, this \\n\"+\n",
       "     \"may be due to a slow or bad network connection. Possible fixes:\\n\"+\n",
       "     \"</p>\\n\"+\n",
       "     \"<ul>\\n\"+\n",
       "     \"<li>re-rerun `output_notebook()` to attempt to load from CDN again, or</li>\\n\"+\n",
       "     \"<li>use INLINE resources instead, as so:</li>\\n\"+\n",
       "     \"</ul>\\n\"+\n",
       "     \"<code>\\n\"+\n",
       "     \"from bokeh.resources import INLINE\\n\"+\n",
       "     \"output_notebook(resources=INLINE)\\n\"+\n",
       "     \"</code>\\n\"+\n",
       "     \"</div>\"}};\n",
       "\n",
       "  function display_loaded() {\n",
       "    var el = document.getElementById(\"1001\");\n",
       "    if (el != null) {\n",
       "      el.textContent = \"BokehJS is loading...\";\n",
       "    }\n",
       "    if (root.Bokeh !== undefined) {\n",
       "      if (el != null) {\n",
       "        el.textContent = \"BokehJS \" + root.Bokeh.version + \" successfully loaded.\";\n",
       "      }\n",
       "    } else if (Date.now() < root._bokeh_timeout) {\n",
       "      setTimeout(display_loaded, 100)\n",
       "    }\n",
       "  }\n",
       "\n",
       "\n",
       "  function run_callbacks() {\n",
       "    try {\n",
       "      root._bokeh_onload_callbacks.forEach(function(callback) {\n",
       "        if (callback != null)\n",
       "          callback();\n",
       "      });\n",
       "    } finally {\n",
       "      delete root._bokeh_onload_callbacks\n",
       "    }\n",
       "    console.debug(\"Bokeh: all callbacks have finished\");\n",
       "  }\n",
       "\n",
       "  function load_libs(css_urls, js_urls, callback) {\n",
       "    if (css_urls == null) css_urls = [];\n",
       "    if (js_urls == null) js_urls = [];\n",
       "\n",
       "    root._bokeh_onload_callbacks.push(callback);\n",
       "    if (root._bokeh_is_loading > 0) {\n",
       "      console.debug(\"Bokeh: BokehJS is being loaded, scheduling callback at\", now());\n",
       "      return null;\n",
       "    }\n",
       "    if (js_urls == null || js_urls.length === 0) {\n",
       "      run_callbacks();\n",
       "      return null;\n",
       "    }\n",
       "    console.debug(\"Bokeh: BokehJS not loaded, scheduling load and callback at\", now());\n",
       "    root._bokeh_is_loading = css_urls.length + js_urls.length;\n",
       "\n",
       "    function on_load() {\n",
       "      root._bokeh_is_loading--;\n",
       "      if (root._bokeh_is_loading === 0) {\n",
       "        console.debug(\"Bokeh: all BokehJS libraries/stylesheets loaded\");\n",
       "        run_callbacks()\n",
       "      }\n",
       "    }\n",
       "\n",
       "    function on_error() {\n",
       "      console.error(\"failed to load \" + url);\n",
       "    }\n",
       "\n",
       "    for (var i = 0; i < css_urls.length; i++) {\n",
       "      var url = css_urls[i];\n",
       "      const element = document.createElement(\"link\");\n",
       "      element.onload = on_load;\n",
       "      element.onerror = on_error;\n",
       "      element.rel = \"stylesheet\";\n",
       "      element.type = \"text/css\";\n",
       "      element.href = url;\n",
       "      console.debug(\"Bokeh: injecting link tag for BokehJS stylesheet: \", url);\n",
       "      document.body.appendChild(element);\n",
       "    }\n",
       "\n",
       "    for (var i = 0; i < js_urls.length; i++) {\n",
       "      var url = js_urls[i];\n",
       "      var element = document.createElement('script');\n",
       "      element.onload = on_load;\n",
       "      element.onerror = on_error;\n",
       "      element.async = false;\n",
       "      element.src = url;\n",
       "      console.debug(\"Bokeh: injecting script tag for BokehJS library: \", url);\n",
       "      document.head.appendChild(element);\n",
       "    }\n",
       "  };var element = document.getElementById(\"1001\");\n",
       "  if (element == null) {\n",
       "    console.error(\"Bokeh: ERROR: autoload.js configured with elementid '1001' but no matching script tag was found. \")\n",
       "    return false;\n",
       "  }\n",
       "\n",
       "  function inject_raw_css(css) {\n",
       "    const element = document.createElement(\"style\");\n",
       "    element.appendChild(document.createTextNode(css));\n",
       "    document.body.appendChild(element);\n",
       "  }\n",
       "\n",
       "  var js_urls = [\"https://cdn.pydata.org/bokeh/release/bokeh-1.3.4.min.js\", \"https://cdn.pydata.org/bokeh/release/bokeh-widgets-1.3.4.min.js\", \"https://cdn.pydata.org/bokeh/release/bokeh-tables-1.3.4.min.js\", \"https://cdn.pydata.org/bokeh/release/bokeh-gl-1.3.4.min.js\"];\n",
       "  var css_urls = [];\n",
       "\n",
       "  var inline_js = [\n",
       "    function(Bokeh) {\n",
       "      Bokeh.set_log_level(\"info\");\n",
       "    },\n",
       "    \n",
       "    function(Bokeh) {\n",
       "      \n",
       "    },\n",
       "    function(Bokeh) {} // ensure no trailing comma for IE\n",
       "  ];\n",
       "\n",
       "  function run_inline_js() {\n",
       "    \n",
       "    if ((root.Bokeh !== undefined) || (force === true)) {\n",
       "      for (var i = 0; i < inline_js.length; i++) {\n",
       "        inline_js[i].call(root, root.Bokeh);\n",
       "      }if (force === true) {\n",
       "        display_loaded();\n",
       "      }} else if (Date.now() < root._bokeh_timeout) {\n",
       "      setTimeout(run_inline_js, 100);\n",
       "    } else if (!root._bokeh_failed_load) {\n",
       "      console.log(\"Bokeh: BokehJS failed to load within specified timeout.\");\n",
       "      root._bokeh_failed_load = true;\n",
       "    } else if (force !== true) {\n",
       "      var cell = $(document.getElementById(\"1001\")).parents('.cell').data().cell;\n",
       "      cell.output_area.append_execute_result(NB_LOAD_WARNING)\n",
       "    }\n",
       "\n",
       "  }\n",
       "\n",
       "  if (root._bokeh_is_loading === 0) {\n",
       "    console.debug(\"Bokeh: BokehJS loaded, going straight to plotting\");\n",
       "    run_inline_js();\n",
       "  } else {\n",
       "    load_libs(css_urls, js_urls, function() {\n",
       "      console.debug(\"Bokeh: BokehJS plotting callback run at\", now());\n",
       "      run_inline_js();\n",
       "    });\n",
       "  }\n",
       "}(window));"
      ],
      "application/vnd.bokehjs_load.v0+json": "\n(function(root) {\n  function now() {\n    return new Date();\n  }\n\n  var force = true;\n\n  if (typeof root._bokeh_onload_callbacks === \"undefined\" || force === true) {\n    root._bokeh_onload_callbacks = [];\n    root._bokeh_is_loading = undefined;\n  }\n\n  \n\n  \n  if (typeof (root._bokeh_timeout) === \"undefined\" || force === true) {\n    root._bokeh_timeout = Date.now() + 5000;\n    root._bokeh_failed_load = false;\n  }\n\n  var NB_LOAD_WARNING = {'data': {'text/html':\n     \"<div style='background-color: #fdd'>\\n\"+\n     \"<p>\\n\"+\n     \"BokehJS does not appear to have successfully loaded. If loading BokehJS from CDN, this \\n\"+\n     \"may be due to a slow or bad network connection. Possible fixes:\\n\"+\n     \"</p>\\n\"+\n     \"<ul>\\n\"+\n     \"<li>re-rerun `output_notebook()` to attempt to load from CDN again, or</li>\\n\"+\n     \"<li>use INLINE resources instead, as so:</li>\\n\"+\n     \"</ul>\\n\"+\n     \"<code>\\n\"+\n     \"from bokeh.resources import INLINE\\n\"+\n     \"output_notebook(resources=INLINE)\\n\"+\n     \"</code>\\n\"+\n     \"</div>\"}};\n\n  function display_loaded() {\n    var el = document.getElementById(\"1001\");\n    if (el != null) {\n      el.textContent = \"BokehJS is loading...\";\n    }\n    if (root.Bokeh !== undefined) {\n      if (el != null) {\n        el.textContent = \"BokehJS \" + root.Bokeh.version + \" successfully loaded.\";\n      }\n    } else if (Date.now() < root._bokeh_timeout) {\n      setTimeout(display_loaded, 100)\n    }\n  }\n\n\n  function run_callbacks() {\n    try {\n      root._bokeh_onload_callbacks.forEach(function(callback) {\n        if (callback != null)\n          callback();\n      });\n    } finally {\n      delete root._bokeh_onload_callbacks\n    }\n    console.debug(\"Bokeh: all callbacks have finished\");\n  }\n\n  function load_libs(css_urls, js_urls, callback) {\n    if (css_urls == null) css_urls = [];\n    if (js_urls == null) js_urls = [];\n\n    root._bokeh_onload_callbacks.push(callback);\n    if (root._bokeh_is_loading > 0) {\n      console.debug(\"Bokeh: BokehJS is being loaded, scheduling callback at\", now());\n      return null;\n    }\n    if (js_urls == null || js_urls.length === 0) {\n      run_callbacks();\n      return null;\n    }\n    console.debug(\"Bokeh: BokehJS not loaded, scheduling load and callback at\", now());\n    root._bokeh_is_loading = css_urls.length + js_urls.length;\n\n    function on_load() {\n      root._bokeh_is_loading--;\n      if (root._bokeh_is_loading === 0) {\n        console.debug(\"Bokeh: all BokehJS libraries/stylesheets loaded\");\n        run_callbacks()\n      }\n    }\n\n    function on_error() {\n      console.error(\"failed to load \" + url);\n    }\n\n    for (var i = 0; i < css_urls.length; i++) {\n      var url = css_urls[i];\n      const element = document.createElement(\"link\");\n      element.onload = on_load;\n      element.onerror = on_error;\n      element.rel = \"stylesheet\";\n      element.type = \"text/css\";\n      element.href = url;\n      console.debug(\"Bokeh: injecting link tag for BokehJS stylesheet: \", url);\n      document.body.appendChild(element);\n    }\n\n    for (var i = 0; i < js_urls.length; i++) {\n      var url = js_urls[i];\n      var element = document.createElement('script');\n      element.onload = on_load;\n      element.onerror = on_error;\n      element.async = false;\n      element.src = url;\n      console.debug(\"Bokeh: injecting script tag for BokehJS library: \", url);\n      document.head.appendChild(element);\n    }\n  };var element = document.getElementById(\"1001\");\n  if (element == null) {\n    console.error(\"Bokeh: ERROR: autoload.js configured with elementid '1001' but no matching script tag was found. \")\n    return false;\n  }\n\n  function inject_raw_css(css) {\n    const element = document.createElement(\"style\");\n    element.appendChild(document.createTextNode(css));\n    document.body.appendChild(element);\n  }\n\n  var js_urls = [\"https://cdn.pydata.org/bokeh/release/bokeh-1.3.4.min.js\", \"https://cdn.pydata.org/bokeh/release/bokeh-widgets-1.3.4.min.js\", \"https://cdn.pydata.org/bokeh/release/bokeh-tables-1.3.4.min.js\", \"https://cdn.pydata.org/bokeh/release/bokeh-gl-1.3.4.min.js\"];\n  var css_urls = [];\n\n  var inline_js = [\n    function(Bokeh) {\n      Bokeh.set_log_level(\"info\");\n    },\n    \n    function(Bokeh) {\n      \n    },\n    function(Bokeh) {} // ensure no trailing comma for IE\n  ];\n\n  function run_inline_js() {\n    \n    if ((root.Bokeh !== undefined) || (force === true)) {\n      for (var i = 0; i < inline_js.length; i++) {\n        inline_js[i].call(root, root.Bokeh);\n      }if (force === true) {\n        display_loaded();\n      }} else if (Date.now() < root._bokeh_timeout) {\n      setTimeout(run_inline_js, 100);\n    } else if (!root._bokeh_failed_load) {\n      console.log(\"Bokeh: BokehJS failed to load within specified timeout.\");\n      root._bokeh_failed_load = true;\n    } else if (force !== true) {\n      var cell = $(document.getElementById(\"1001\")).parents('.cell').data().cell;\n      cell.output_area.append_execute_result(NB_LOAD_WARNING)\n    }\n\n  }\n\n  if (root._bokeh_is_loading === 0) {\n    console.debug(\"Bokeh: BokehJS loaded, going straight to plotting\");\n    run_inline_js();\n  } else {\n    load_libs(css_urls, js_urls, function() {\n      console.debug(\"Bokeh: BokehJS plotting callback run at\", now());\n      run_inline_js();\n    });\n  }\n}(window));"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.metrics import confusion_matrix\n",
    "from bokeh.io import show, output_notebook\n",
    "from bokeh.layouts import row\n",
    "from bokeh.plotting import figure\n",
    "output_notebook()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels = [\"😂\", \"😊\", \"😒\", \"😭\"] \n",
    "pred = model_1.predict(valid_data)\n",
    "y_true = np.argmax(valid_label, axis = 1)\n",
    "y_pred = np.argmax(pred, axis = 1)\n",
    "confusion_matrix_result = confusion_matrix(y_true, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_confusion_matrix(labels, confusion_matrix_result):\n",
    "    print(\"Confusion matrix:\")\n",
    "    print(confusion_matrix_result)\n",
    "    num_max = np.max(confusion_matrix_result)\n",
    "    #num_min = np.min(confusion_matrix_result)\n",
    "    num_min = 0\n",
    "    xname = []\n",
    "    yname = []\n",
    "    color = []\n",
    "    alpha = []\n",
    "    count = []\n",
    "\n",
    "    for i, label_1 in enumerate(labels):\n",
    "        for j, label_2 in enumerate(labels):\n",
    "            xname.append(label_1)\n",
    "            yname.append(label_2)\n",
    "            color.append('#008000')\n",
    "            alpha.append((confusion_matrix_result[j, i] - num_min) / (num_max - num_min))\n",
    "            count.append(confusion_matrix_result[j, i])\n",
    "\n",
    "    data = dict(\n",
    "        xname = xname,\n",
    "        yname = yname,\n",
    "        colors = color,\n",
    "        alphas = alpha,\n",
    "        count = count,\n",
    "    )\n",
    "\n",
    "    p = figure(title = \"Confusion Matrix\",\n",
    "               x_axis_location = \"below\", tools = \"hover, save\",\n",
    "               x_range = labels, \n",
    "               y_range=list(reversed(labels)),\n",
    "               tooltips = [('True Label', '@yname'), \n",
    "                           ('Predicted Label', '@xname'),\n",
    "                           ('Count', '@count')])\n",
    "\n",
    "    p.plot_width = 400\n",
    "    p.plot_height = 400\n",
    "    p.grid.grid_line_color = None\n",
    "    p.axis.axis_line_color = None\n",
    "    p.axis.major_tick_line_color = None\n",
    "    p.axis.major_label_text_font_size = \"15pt\"\n",
    "    p.axis.major_label_standoff = 0\n",
    "    p.xaxis.axis_label = 'Predicted Label'\n",
    "    p.yaxis.axis_label = 'True Label'\n",
    "    \n",
    "    p.rect('xname', 'yname', 0.9, 0.9, source = data,\n",
    "           color =  'colors', alpha = 'alphas', line_color = None,\n",
    "           hover_line_color = 'black', hover_color = 'colors', \n",
    "           hover_alpha = 'alphas')\n",
    "\n",
    "    show(p, notebook_handle= True) # show the plot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion matrix:\n",
      "[[1652  482  452  569]\n",
      " [ 511 2513  156  301]\n",
      " [ 602  196 1505  546]\n",
      " [ 702  418  521 2087]]\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "\n",
       "\n",
       "\n",
       "\n",
       "\n",
       "  <div class=\"bk-root\" id=\"ea11300b-2ff9-4f46-9591-510ce1090144\" data-root-id=\"1002\"></div>\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/javascript": [
       "(function(root) {\n",
       "  function embed_document(root) {\n",
       "    \n",
       "  var docs_json = {\"b0ba4a63-a421-4ac2-bef1-552d95aa4a31\":{\"roots\":{\"references\":[{\"attributes\":{\"below\":[{\"id\":\"1013\",\"type\":\"CategoricalAxis\"}],\"center\":[{\"id\":\"1016\",\"type\":\"Grid\"},{\"id\":\"1020\",\"type\":\"Grid\"}],\"left\":[{\"id\":\"1017\",\"type\":\"CategoricalAxis\"}],\"plot_height\":400,\"plot_width\":400,\"renderers\":[{\"id\":\"1031\",\"type\":\"GlyphRenderer\"}],\"title\":{\"id\":\"1003\",\"type\":\"Title\"},\"toolbar\":{\"id\":\"1023\",\"type\":\"Toolbar\"},\"x_range\":{\"id\":\"1005\",\"type\":\"FactorRange\"},\"x_scale\":{\"id\":\"1009\",\"type\":\"CategoricalScale\"},\"y_range\":{\"id\":\"1007\",\"type\":\"FactorRange\"},\"y_scale\":{\"id\":\"1011\",\"type\":\"CategoricalScale\"}},\"id\":\"1002\",\"subtype\":\"Figure\",\"type\":\"Plot\"},{\"attributes\":{\"fill_alpha\":{\"value\":0.1},\"fill_color\":{\"value\":\"#1f77b4\"},\"height\":{\"units\":\"data\",\"value\":0.9},\"line_alpha\":{\"value\":0.1},\"line_color\":{\"value\":\"#1f77b4\"},\"width\":{\"units\":\"data\",\"value\":0.9},\"x\":{\"field\":\"xname\"},\"y\":{\"field\":\"yname\"}},\"id\":\"1029\",\"type\":\"Rect\"},{\"attributes\":{\"callback\":null,\"factors\":[\"\\ud83d\\ude2d\",\"\\ud83d\\ude12\",\"\\ud83d\\ude0a\",\"\\ud83d\\ude02\"]},\"id\":\"1007\",\"type\":\"FactorRange\"},{\"attributes\":{},\"id\":\"1011\",\"type\":\"CategoricalScale\"},{\"attributes\":{\"axis_label\":\"Predicted Label\",\"axis_line_color\":{\"value\":null},\"formatter\":{\"id\":\"1035\",\"type\":\"CategoricalTickFormatter\"},\"major_label_standoff\":0,\"major_label_text_font_size\":{\"value\":\"15pt\"},\"major_tick_line_color\":{\"value\":null},\"ticker\":{\"id\":\"1014\",\"type\":\"CategoricalTicker\"}},\"id\":\"1013\",\"type\":\"CategoricalAxis\"},{\"attributes\":{\"fill_alpha\":{\"field\":\"alphas\"},\"fill_color\":{\"field\":\"colors\"},\"height\":{\"units\":\"data\",\"value\":0.9},\"line_alpha\":{\"field\":\"alphas\"},\"width\":{\"units\":\"data\",\"value\":0.9},\"x\":{\"field\":\"xname\"},\"y\":{\"field\":\"yname\"}},\"id\":\"1030\",\"type\":\"Rect\"},{\"attributes\":{},\"id\":\"1022\",\"type\":\"SaveTool\"},{\"attributes\":{\"text\":\"Confusion Matrix\"},\"id\":\"1003\",\"type\":\"Title\"},{\"attributes\":{\"callback\":null,\"tooltips\":[[\"True Label\",\"@yname\"],[\"Predicted Label\",\"@xname\"],[\"Count\",\"@count\"]]},\"id\":\"1021\",\"type\":\"HoverTool\"},{\"attributes\":{},\"id\":\"1018\",\"type\":\"CategoricalTicker\"},{\"attributes\":{},\"id\":\"1039\",\"type\":\"Selection\"},{\"attributes\":{\"data_source\":{\"id\":\"1026\",\"type\":\"ColumnDataSource\"},\"glyph\":{\"id\":\"1028\",\"type\":\"Rect\"},\"hover_glyph\":{\"id\":\"1030\",\"type\":\"Rect\"},\"muted_glyph\":null,\"nonselection_glyph\":{\"id\":\"1029\",\"type\":\"Rect\"},\"selection_glyph\":null,\"view\":{\"id\":\"1032\",\"type\":\"CDSView\"}},\"id\":\"1031\",\"type\":\"GlyphRenderer\"},{\"attributes\":{},\"id\":\"1035\",\"type\":\"CategoricalTickFormatter\"},{\"attributes\":{\"fill_alpha\":{\"field\":\"alphas\"},\"fill_color\":{\"field\":\"colors\"},\"height\":{\"units\":\"data\",\"value\":0.9},\"line_alpha\":{\"field\":\"alphas\"},\"line_color\":{\"value\":null},\"width\":{\"units\":\"data\",\"value\":0.9},\"x\":{\"field\":\"xname\"},\"y\":{\"field\":\"yname\"}},\"id\":\"1028\",\"type\":\"Rect\"},{\"attributes\":{\"active_drag\":\"auto\",\"active_inspect\":\"auto\",\"active_multi\":null,\"active_scroll\":\"auto\",\"active_tap\":\"auto\",\"tools\":[{\"id\":\"1021\",\"type\":\"HoverTool\"},{\"id\":\"1022\",\"type\":\"SaveTool\"}]},\"id\":\"1023\",\"type\":\"Toolbar\"},{\"attributes\":{},\"id\":\"1037\",\"type\":\"CategoricalTickFormatter\"},{\"attributes\":{},\"id\":\"1038\",\"type\":\"UnionRenderers\"},{\"attributes\":{\"callback\":null,\"data\":{\"alphas\":[0.6573816155988857,0.20334261838440112,0.2395543175487465,0.2793473935535217,0.1918026263430163,1.0,0.07799442896935933,0.1663350576999602,0.17986470354158376,0.06207719856744926,0.5988857938718662,0.20732192598487864,0.22642260246717072,0.11977715877437325,0.21727019498607242,0.8304814962196578],\"colors\":[\"#008000\",\"#008000\",\"#008000\",\"#008000\",\"#008000\",\"#008000\",\"#008000\",\"#008000\",\"#008000\",\"#008000\",\"#008000\",\"#008000\",\"#008000\",\"#008000\",\"#008000\",\"#008000\"],\"count\":[1652,511,602,702,482,2513,196,418,452,156,1505,521,569,301,546,2087],\"xname\":[\"\\ud83d\\ude02\",\"\\ud83d\\ude02\",\"\\ud83d\\ude02\",\"\\ud83d\\ude02\",\"\\ud83d\\ude0a\",\"\\ud83d\\ude0a\",\"\\ud83d\\ude0a\",\"\\ud83d\\ude0a\",\"\\ud83d\\ude12\",\"\\ud83d\\ude12\",\"\\ud83d\\ude12\",\"\\ud83d\\ude12\",\"\\ud83d\\ude2d\",\"\\ud83d\\ude2d\",\"\\ud83d\\ude2d\",\"\\ud83d\\ude2d\"],\"yname\":[\"\\ud83d\\ude02\",\"\\ud83d\\ude0a\",\"\\ud83d\\ude12\",\"\\ud83d\\ude2d\",\"\\ud83d\\ude02\",\"\\ud83d\\ude0a\",\"\\ud83d\\ude12\",\"\\ud83d\\ude2d\",\"\\ud83d\\ude02\",\"\\ud83d\\ude0a\",\"\\ud83d\\ude12\",\"\\ud83d\\ude2d\",\"\\ud83d\\ude02\",\"\\ud83d\\ude0a\",\"\\ud83d\\ude12\",\"\\ud83d\\ude2d\"]},\"selected\":{\"id\":\"1039\",\"type\":\"Selection\"},\"selection_policy\":{\"id\":\"1038\",\"type\":\"UnionRenderers\"}},\"id\":\"1026\",\"type\":\"ColumnDataSource\"},{\"attributes\":{\"dimension\":1,\"grid_line_color\":null,\"ticker\":{\"id\":\"1018\",\"type\":\"CategoricalTicker\"}},\"id\":\"1020\",\"type\":\"Grid\"},{\"attributes\":{\"axis_label\":\"True Label\",\"axis_line_color\":{\"value\":null},\"formatter\":{\"id\":\"1037\",\"type\":\"CategoricalTickFormatter\"},\"major_label_standoff\":0,\"major_label_text_font_size\":{\"value\":\"15pt\"},\"major_tick_line_color\":{\"value\":null},\"ticker\":{\"id\":\"1018\",\"type\":\"CategoricalTicker\"}},\"id\":\"1017\",\"type\":\"CategoricalAxis\"},{\"attributes\":{\"source\":{\"id\":\"1026\",\"type\":\"ColumnDataSource\"}},\"id\":\"1032\",\"type\":\"CDSView\"},{\"attributes\":{},\"id\":\"1009\",\"type\":\"CategoricalScale\"},{\"attributes\":{\"grid_line_color\":null,\"ticker\":{\"id\":\"1014\",\"type\":\"CategoricalTicker\"}},\"id\":\"1016\",\"type\":\"Grid\"},{\"attributes\":{\"callback\":null,\"factors\":[\"\\ud83d\\ude02\",\"\\ud83d\\ude0a\",\"\\ud83d\\ude12\",\"\\ud83d\\ude2d\"]},\"id\":\"1005\",\"type\":\"FactorRange\"},{\"attributes\":{},\"id\":\"1014\",\"type\":\"CategoricalTicker\"}],\"root_ids\":[\"1002\"]},\"title\":\"Bokeh Application\",\"version\":\"1.3.4\"}};\n",
       "  var render_items = [{\"docid\":\"b0ba4a63-a421-4ac2-bef1-552d95aa4a31\",\"notebook_comms_target\":\"1040\",\"roots\":{\"1002\":\"ea11300b-2ff9-4f46-9591-510ce1090144\"}}];\n",
       "  root.Bokeh.embed.embed_items_notebook(docs_json, render_items);\n",
       "\n",
       "  }\n",
       "  if (root.Bokeh !== undefined) {\n",
       "    embed_document(root);\n",
       "  } else {\n",
       "    var attempts = 0;\n",
       "    var timer = setInterval(function(root) {\n",
       "      if (root.Bokeh !== undefined) {\n",
       "        embed_document(root);\n",
       "        clearInterval(timer);\n",
       "      }\n",
       "      attempts++;\n",
       "      if (attempts > 100) {\n",
       "        console.log(\"Bokeh: ERROR: Unable to run BokehJS code because BokehJS library is missing\");\n",
       "        clearInterval(timer);\n",
       "      }\n",
       "    }, 10, root)\n",
       "  }\n",
       "})(window);"
      ],
      "application/vnd.bokehjs_exec.v0+json": ""
     },
     "metadata": {
      "application/vnd.bokehjs_exec.v0+json": {
       "id": "1002"
      }
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_confusion_matrix(labels, confusion_matrix_result)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Show mislabelled examples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5456 out of 13213 tweets are mispredicted in validation set\n"
     ]
    }
   ],
   "source": [
    "valid_index = np.load(\"data/valid_index.npy\")\n",
    "predict_index = np.concatenate((valid_index, y_pred.reshape(y_pred.shape[0], 1)), axis = 1)\n",
    "miss_predict_index = predict_index[y_true != y_pred]\n",
    "print(\"{} out of {} tweets are mispredicted in validation set\".format(miss_predict_index.shape[0], \n",
    "                                                                      predict_index.shape[0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_0 = pd.read_csv(\"data/clean/positive.csv\", encoding= \"utf-8\")\n",
    "df_1 = pd.read_csv(\"data/clean/slightly_positive.csv\", encoding= \"utf-8\")\n",
    "df_2 = pd.read_csv(\"data/clean/slightly_negative.csv\", encoding= \"utf-8\")\n",
    "df_3 = pd.read_csv(\"data/clean/negative.csv\", encoding= \"utf-8\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [],
   "source": [
    "def show_example(index_list, labels):\n",
    "    idx = index_list[0]\n",
    "    label_true = index_list[1]\n",
    "    label_mispredict = index_list[2]\n",
    "    if label_true == 0:\n",
    "        text = df_0.text[idx]\n",
    "    elif label_true == 1:\n",
    "        text = df_1.text[idx]\n",
    "    elif label_true == 2:\n",
    "        text = df_2.text[idx]\n",
    "    elif label_true == 3:\n",
    "        text = df_3.text[idx]\n",
    "    print(\"Predict: {}\\tTrue: {}\\nText: {}\".format(labels[label_mispredict], labels[label_true], text))\n",
    "    print(\"---------------------------------------------------------\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Predict: 😭\tTrue: 😒\n",
      "Text: i downloaded her album, and got bored ndingekafiki naku track 5 😒\n",
      "---------------------------------------------------------\n",
      "Predict: 😂\tTrue: 😭\n",
      "Text: @arubaology yes yall do djdjdjdjdj the us aau teams stay playing canadian ones 😭\n",
      "---------------------------------------------------------\n",
      "Predict: 😒\tTrue: 😂\n",
      "Text: Nonsense.....  kwankara is a tribal ethnic group\n",
      "does it put ants in your pants? 😂 https://t.co/wGXLeS1fAL\n",
      "---------------------------------------------------------\n",
      "Predict: 😂\tTrue: 😊\n",
      "Text: rochelleniehues was doing a Guessing Game And I Won 😊 https://t.co/cchGJm9DFj\n",
      "---------------------------------------------------------\n",
      "Predict: 😭\tTrue: 😒\n",
      "Text: my ppls have it hot asf in here ..let me turn my air on 😒\n",
      "---------------------------------------------------------\n",
      "Predict: 😭\tTrue: 😂\n",
      "Text: Idc if us bitches be roastin each other I️ just appreciate that they let each other finish 😂 https://t.co/LsvkZHm7V8\n",
      "---------------------------------------------------------\n",
      "Predict: 😭\tTrue: 😂\n",
      "Text: on god I️ was high asf tryna take care of yo ass 😂 https://t.co/CHwUKQdqZk\n",
      "---------------------------------------------------------\n",
      "Predict: 😭\tTrue: 😒\n",
      "Text: i really don’t wanna work tomorrow 😒\n",
      "---------------------------------------------------------\n",
      "Predict: 😂\tTrue: 😭\n",
      "Text: It really be the niggas. 😭 https://t.co/YIWcELUOE8\n",
      "---------------------------------------------------------\n",
      "Predict: 😂\tTrue: 😊\n",
      "Text: This guy 😊 he's witty, super smart, tells great dad jokes, is so cool that this is one of the… https://t.co/HoaF4l0bI0\n",
      "---------------------------------------------------------\n",
      "Predict: 😭\tTrue: 😂\n",
      "Text: @foolishfoolie Lmao please stop this 😂\n",
      "---------------------------------------------------------\n",
      "Predict: 😒\tTrue: 😂\n",
      "Text: @TheRealBozza @TheRunHome 3-1 is a loss. Even if you give an a away goal as 2, it's still a defeat 😂\n",
      "---------------------------------------------------------\n",
      "Predict: 😭\tTrue: 😂\n",
      "Text: @MannyMua733 I mean, it's fucked up, but I still want it 😂\n",
      "---------------------------------------------------------\n",
      "Predict: 😒\tTrue: 😭\n",
      "Text: @biebsshawty99 @SelenaFanClub it's always USA and London, ew. 😭\n",
      "---------------------------------------------------------\n",
      "Predict: 😭\tTrue: 😒\n",
      "Text: @breakkingground I️ wish I️ was. I️ have class 😒\n",
      "---------------------------------------------------------\n",
      "Predict: 😊\tTrue: 😭\n",
      "Text: @DIYSOS Every time! 😭 So inspirational. Well Done!!\n",
      "---------------------------------------------------------\n",
      "Predict: 😂\tTrue: 😒\n",
      "Text: #Repost lyndlkean\n",
      "・・・\n",
      "Are you talking to me? 😒 https://t.co/fvogzxmzf7\n",
      "---------------------------------------------------------\n",
      "Predict: 😊\tTrue: 😂\n",
      "Text: @realjasonsmash 😂 thanks my boy!\n",
      "---------------------------------------------------------\n",
      "Predict: 😭\tTrue: 😂\n",
      "Text: Y’all I️ had to pee so bad I️ almost peed on myself ! 😂\n",
      "---------------------------------------------------------\n",
      "Predict: 😂\tTrue: 😒\n",
      "Text: @philipoconnor Remember the time brendan said he thought preteen girls going to concert looked like sex workers. 😒 Sound.\n",
      "---------------------------------------------------------\n",
      "Predict: 😂\tTrue: 😒\n",
      "Text: @blushyouth What the hahahaha stop being a stingy girl ji 😒 how about New York's Plastic bag? Hahaha\n",
      "---------------------------------------------------------\n",
      "Predict: 😂\tTrue: 😒\n",
      "Text: @IndieWire He flashed his cock? 🙈 Oh my, lock him up and throw away the key! 😒 #Lame\n",
      "---------------------------------------------------------\n",
      "Predict: 😊\tTrue: 😒\n",
      "Text: Who’s still up? 😒\n",
      "---------------------------------------------------------\n",
      "Predict: 😒\tTrue: 😂\n",
      "Text: Trying to be laid up smoking a blunt with somebody son 😂\n",
      "---------------------------------------------------------\n",
      "Predict: 😊\tTrue: 😭\n",
      "Text: I had completely forgotten this drawing even existed but on the bright side it's going to be pretty easy to redraw it hahaha... 😭\n",
      "---------------------------------------------------------\n",
      "Predict: 😒\tTrue: 😭\n",
      "Text: Literally lost ALL my contacts 😭 like wtf 😩\n",
      "---------------------------------------------------------\n",
      "Predict: 😊\tTrue: 😂\n",
      "Text: @CLOgilvie @natornatalie 😂 love this!\n",
      "---------------------------------------------------------\n",
      "Predict: 😒\tTrue: 😭\n",
      "Text: If she is not wet it's your fault, if you can't get hard it's still your fault again 😭 https://t.co/nm5WeyD8iy\n",
      "---------------------------------------------------------\n",
      "Predict: 😂\tTrue: 😭\n",
      "Text: @sugaplumcris coming from you man 😭\n",
      "---------------------------------------------------------\n",
      "Predict: 😂\tTrue: 😭\n",
      "Text: @Masterofnone___ Girls say anything nowadays 😭\n",
      "---------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "for i in np.random.choice(len(miss_predict_index), 30, replace=False):\n",
    "    show_example(miss_predict_index[i], labels)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python (ML)",
   "language": "python",
   "name": "ml"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}