Part III - Training and Development of Different LSTM Models.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "bK4FOeG5ZtWs"
   },
   "source": [
    "<h1 align=\"center\">PART III</h1>\n",
    "<h1 align=\"center\">Training and Development of Different LSTM Models</h1> "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "RK0igzi1ZtWu"
   },
   "source": [
    "To get started, let's download the necessary libraries:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "id": "207XAbAlZtWv"
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import pickle\n",
    "import gensim, logging\n",
    "import gensim.models.keyedvectors as word2vec\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "id": "ahbTiUEMqzkv"
   },
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import LSTM\n",
    "from keras.layers.core import Dense\n",
    "from keras.models import load_model\n",
    "from keras.layers.core import Dropout"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "T0KKdQILZtWw"
   },
   "source": [
    "Also let's write a style for alignment in the middle of all graphs, images, etc:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 17
    },
    "id": "9qrGx_YKZtWx",
    "outputId": "37a21809-3dca-4565-f851-f89e2b4062da"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "<style>\n",
       ".output_png {\n",
       "    display: table-cell;\n",
       "    text-align: center;\n",
       "    vertical-align: middle;\n",
       "}\n",
       "</style>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from IPython.core.display import HTML\n",
    "HTML(\"\"\"\n",
    "<style>\n",
    ".output_png {\n",
    "    display: table-cell;\n",
    "    text-align: center;\n",
    "    vertical-align: middle;\n",
    "}\n",
    "</style>\n",
    "\"\"\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "TmKwMBq8ZtWy"
   },
   "source": [
    "<h2 align=\"center\">Preparation of input data</h2> \n",
    "\n",
    "Firstly, we will load the sample data we processed in the first part:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "id": "SF3wgaQgZtWy"
   },
   "outputs": [],
   "source": [
    "with open('documents.pql', 'rb') as f:\n",
    "     docs = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "id": "2FZkuKUSyd0Z"
   },
   "outputs": [],
   "source": [
    "import random\n",
    "\n",
    "pos = [doc for doc in docs if doc[1] == \"pos\"][:15000]\n",
    "neg = [doc for doc in docs if doc[1] == \"neg\"][:15000]\n",
    "\n",
    "docs = pos + neg\n",
    "random.shuffle(docs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "7DC5weErZtWz",
    "outputId": "e07b634e-07d9-4857-e42a-98eb178b980a"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of documents: 30000\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of documents:\", len(docs))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "5RLxmbUBZtWz"
   },
   "source": [
    "Google Word2Vec model is very difficult to process with my RAM, so we will use a smaller model, namely GloVe, model contains 400000 words, so we will get a 400000 x 100 embedding matrix that contains all the values of the word vectors.\n",
    "\n",
    "You can download model [here](https://www.kaggle.com/datasets/anindya2906/glove6b?select=glove.6B.100d.txt)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "id": "LVW8xdVRZtWz"
   },
   "outputs": [],
   "source": [
    "model = word2vec.KeyedVectors.load_word2vec_format('glove.6B.100d.w2vformat.txt', binary=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "A1BFwzhWZtW0"
   },
   "source": [
    "Now let's get a list of all the words from our dictionary:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "id": "GOdz3u48ZtW0"
   },
   "outputs": [],
   "source": [
    "words = list(model.index_to_key)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "V0z2NooeZtW0"
   },
   "source": [
    "Just to make sure everything is loaded correctly, we can look at the dimensions of the dictionary list and the embedding matrix:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "5t5MEbssZtW0",
    "outputId": "14b9e244-0228-4706-e4c2-9df21eae5178"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total words: 400000 \n",
      "\n",
      "Word-Vectors shape: (400000, 100)\n"
     ]
    }
   ],
   "source": [
    "print(\"Total words:\", len(words), \"\\n\\nWord-Vectors shape:\", model.vectors.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "CSQ4xxThZtW1"
   },
   "source": [
    "The shape of our sample for training will be `(H, N, 100)`, where:\n",
    "* `H` - number of samples;\n",
    "* `N` - number of words in each sentence;\n",
    "* `100` - dimension of each word.\n",
    "\n",
    "Let's analyze how many words are usually found in the reviews:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "m1IYxmuPZtW1",
    "outputId": "4202c7d4-a762-4551-9c44-926e0b74d95e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average number of words in the review: 46\n",
      "\n",
      "Maximum review length: 2376 \n",
      "Minimum review length: 1\n",
      "\n",
      "Review with more than 50 words: 8319 \n",
      "Review with more than 100 words: 2790 \n",
      "Review with more than 200 words: 693 \n",
      "Review with more than 300 words: 225 \n",
      "Review with more than 400 words: 98\n"
     ]
    }
   ],
   "source": [
    "max_word, min_word = 0, 100\n",
    "word50, word100, word200, word300, word400  = 0, 0, 0, 0, 0\n",
    "reviews_len = []\n",
    "\n",
    "for review, state in docs:\n",
    "    reviews_len.append(len(review))\n",
    "\n",
    "    if len(review) > max_word: max_word = len(review)\n",
    "    if len(review) < min_word: min_word = len(review)\n",
    "\n",
    "    if len(review) > 50: word50 += 1\n",
    "    if len(review) > 100: word100 += 1\n",
    "    if len(review) > 200: word200 += 1\n",
    "    if len(review) > 300: word300 += 1\n",
    "    if len(review) > 400: word400 += 1\n",
    "    \n",
    "print(\"Average number of words in the review:\", int(sum(reviews_len)/len(reviews_len)))\n",
    "print(\"\\nMaximum review length:\", max_word, \"\\nMinimum review length:\", min_word)\n",
    "print(\"\\nReview with more than 50 words:\", word50, \n",
    "      \"\\nReview with more than 100 words:\", word100,\n",
    "      \"\\nReview with more than 200 words:\", word200, \n",
    "      \"\\nReview with more than 300 words:\", word300,\n",
    "      \"\\nReview with more than 400 words:\", word400)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "7iaKaR1BZtW1"
   },
   "source": [
    "Visualize the frequency of words in the review:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 295
    },
    "id": "Pk2asIOMZtW2",
    "outputId": "604ffe6b-3387-49cf-fb87-7951589e36f9"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(reviews_len, bins=150)\n",
    "plt.axis([0, 400, 0, 10000])\n",
    "plt.ylabel('Frequency')\n",
    "plt.xlabel('Sequence Length');\n",
    "plt.title(\"Average number of words\");"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "NwKoHdVvZtW2"
   },
   "source": [
    "Our next function `fix_review_len()` will do a useful job, this function is designed to fix the size of reviews to a fixed size to feed them into a neural network with reviews of a certain length. Reviews whose length is less than fixed will be extended by zeros. This process does not affect the algorithm and reviews longer than the specified length will be truncated."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "wbTlHCmAZtW2",
    "outputId": "088d096f-af65-4838-a812-bd4269da19f8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[3, 1, 2, 4, 5, [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]\n"
     ]
    }
   ],
   "source": [
    "def fix_review_len(review, length):\n",
    "    if len(review) > length:\n",
    "        review = review[:length]\n",
    "    elif len(review) < length:\n",
    "        for i in range(length - len(review)):\n",
    "            zeros = [0] * 100\n",
    "            review.append(zeros)\n",
    "    return review\n",
    "\n",
    "example = [3, 1, 2, 4, 5]\n",
    "example = fix_review_len(example, 7)\n",
    "print(example) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "onLNFH-bZtW3"
   },
   "source": [
    "Now let's slightly improve our `sent_embed()` function from previous part and then update our dataset with word vectors.\n",
    "\n",
    "Based on the histogram data, as well as the average number of words in the files, we can say with confidence that most reviews will have less than 100 words, which is the maximum value of the length of the sequence that we set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "id": "h58uiBG8ZtW3"
   },
   "outputs": [],
   "source": [
    "def sent_embed(words, docs):\n",
    "    x_sent_embed, y_sent_embed = [], [] \n",
    "    \n",
    "    max_seq_len = 100\n",
    "    \n",
    "    # recover the embedding of each sentence with the average of the vector that composes it\n",
    "    # sent - sentence, state - state of the sentence (pos/neg)\n",
    "    for sent, state in docs:\n",
    "        # average embedding of all words in a sentence\n",
    "        sent_embed = []\n",
    "        for word in sent:\n",
    "            try:\n",
    "                # if word is present in the dictionary - add its vector representation\n",
    "                sent_embed.append(model[word])\n",
    "            except KeyError:\n",
    "                # if word is not in the dictionary - add a zero vector\n",
    "                sent_embed.append([0] * 100)\n",
    "        \n",
    "        # add a sentence vector to the list\n",
    "        sent_embed = fix_review_len(sent_embed, max_seq_len)\n",
    "        x_sent_embed.append(sent_embed)\n",
    "        \n",
    "        # add a label to y_sent_embed\n",
    "        if state == 'pos': y_sent_embed.append(1)\n",
    "        elif state == 'neg': y_sent_embed.append(0)\n",
    "            \n",
    "    return x_sent_embed, y_sent_embed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "id": "2BePmRm3ZtW4"
   },
   "outputs": [],
   "source": [
    "x, y = sent_embed(words, docs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "dALGpRR9ZtW4",
    "outputId": "0bac5394-0351-42bb-d333-486ab521c4fc"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of X: (30000, 100, 100)\n"
     ]
    }
   ],
   "source": [
    "x = np.array(x)\n",
    "print(\"Shape of X:\", x.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "b-5DDr5wZtW4"
   },
   "source": [
    "<h3 align=\"center\">Split Corpus</h3>\n",
    "\n",
    "Now, for further work, we will divide our corpus for training, testing and development sets:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "id": "9vlTQXzAZtW4"
   },
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# train test\n",
    "x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.15, random_state=42)\n",
    "\n",
    "# train dev\n",
    "x_train, x_val, y_train, y_val = train_test_split(x_train, y_train, test_size=0.15, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "gal_WGq-ZtW5",
    "outputId": "109d67b2-df23-4519-e48f-6d450c9ce0a5"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Length of x_train: 21675 | Length of y_train: 21675\n",
      "Length of x_test:   4500 | Length of y_test:  4500\n",
      "Length of x_val:    3825 | Length of y_val:   3825\n"
     ]
    }
   ],
   "source": [
    "print('Length of x_train:', len(x_train), '| Length of y_train:', len(y_train))\n",
    "print('Length of x_test:  ', len(x_test), '| Length of y_test: ', len(y_test))\n",
    "print('Length of x_val:   ', len(x_val), '| Length of y_val:  ', len(y_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "Gk96bR9OZtW5",
    "outputId": "f584b82c-a5a2-4e04-8110-03cfeebb6f3f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of x_train set: (21675, 100, 100)\n"
     ]
    }
   ],
   "source": [
    "print(\"Shape of x_train set:\", x_train.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "UO05ZRsXZtW5"
   },
   "source": [
    "<h2 align=\"center\">LSTM Model (Stochastic Gradient Descent)</h2>\n",
    "\n",
    "Keras's benefit is that it is built on top of symbolic mathematical libraries such as TensorFlow and Theano for fast and efficient computation. This is needed with large neural networks.\n",
    "\n",
    "A downside of using these efficient libraries is that you must define the scope of your data upfront and for all time. Specifically, the batch size.\n",
    "\n",
    "Neural networks are trained using the stochastic gradient descent optimization algorithm.\n",
    "\n",
    "This involves using the current state of the model to make a prediction, comparing the prediction to the expected values, and using the difference as an estimate of the error gradient. This error gradient is then used to update the model weights and the process is repeated.\n",
    "\n",
    "The error gradient is a statistical estimate. The more training examples used in the estimate, the more accurate this estimate will be and the more likely that the weights of the network will be adjusted in a way that will improve the performance of the model. The improved estimate of the error gradient comes at the cost of having to use the model to make many more predictions before the estimate can be calculated, and in turn, the weights updated.\n",
    "\n",
    "Alternately, using fewer examples results in a less accurate estimate of the error gradient that is highly dependent on the specific training examples used.\n",
    "\n",
    "This results in a noisy estimate that, in turn, results in noisy updates to the model weights, e.g. many updates with perhaps quite different estimates of the error gradient. Nevertheless, these noisy updates can result in faster learning and sometimes a more robust model.\n",
    "\n",
    "The number of training examples used in the estimate of the error gradient is a hyperparameter for the learning algorithm called the “batch size,” or simply the “batch.”\n",
    "\n",
    "A batch size of 32 means that 32 samples from the training dataset will be used to estimate the error gradient before the model weights are updated. One training epoch means that the learning algorithm has made one pass through the training dataset, where examples were separated into randomly selected “batch size” groups.\n",
    "\n",
    "Historically, a training algorithm where the batch size is set to the total number of training examples is called “batch gradient descent” and a training algorithm where the batch size is set to 1 training example is called “stochastic gradient descent” or “online gradient descent.”\n",
    "\n",
    "A configuration of the batch size anywhere in between (e.g. more than 1 example and less than the number of examples in the training dataset) is called “minibatch gradient descent.”\n",
    "\n",
    "  - **Batch Gradient Descent**. Batch size is set to the total number of examples in the training dataset.\n",
    "  - **Stochastic Gradient Descent**. Batch size is set to one.\n",
    "  - **Minibatch Gradient Descent**. Batch size is set to more than one and less than the total number of examples in the training dataset.\n",
    "For shorthand, the algorithm is often referred to as stochastic gradient descent regardless of the batch size. Given that very large datasets are often used to train deep learning neural networks, the batch size is rarely set to the size of the training dataset.\n",
    "\n",
    "Smaller batch sizes are used for two main reasons:\n",
    "\n",
    "  - Smaller batch sizes are noisy, offering a regularizing effect and lower generalization error.\n",
    "  - Smaller batch sizes make it easier to fit one batch worth of training data in memory (i.e. when using a GPU).\n",
    "A third reason is that the batch size is often set at something small, such as 32 examples, and is not tuned by the practitioner. Small batch sizes such as 32 do work well generally.\n",
    "\n",
    "Nevertheless, the batch size impacts how quickly a model learns and the stability of the learning process. It is an important hyperparameter that should be well understood and tuned by the deep learning practitioner."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "RsfoHjRbZtW5"
   },
   "source": [
    "<h3 align=\"center\">Setting Up and Creating the Network</h3>\n",
    "\n",
    "The network has one input, a hidden layer with 128 units, and an output layer with 1 unit.\n",
    "\n",
    "A mean squared error optimization function is used for this problem with the efficient ADAM optimization algorithm.\n",
    "\n",
    "So now let's configure and create the network:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "lMn08CvVZtW6",
    "outputId": "8275bc71-7f80-4cf2-97f7-a176c5cedd77"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " lstm (LSTM)                 (None, 128)               117248    \n",
      "                                                                 \n",
      " dense (Dense)               (None, 1)                 129       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 117,377\n",
      "Trainable params: 117,377\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "# configure network\n",
    "n_neurons = 128\n",
    "n_epoch = 20\n",
    "n_batch = 1\n",
    "\n",
    "# design network\n",
    "model_batch = Sequential()\n",
    "model_batch.add(LSTM(n_neurons, input_shape=(x.shape[1], x.shape[2])))\n",
    "model_batch.add(Dense(1, activation='sigmoid'))\n",
    "model_batch.compile(loss='mean_squared_error', optimizer='adam', metrics=['binary_accuracy'])\n",
    "\n",
    "print(model_batch.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "P_Jz8mfqZtW6",
    "outputId": "7147bdc9-9aa5-443d-d05c-f09f718bfc4c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "21675/21675 [==============================] - 814s 37ms/step - loss: 0.2422 - binary_accuracy: 0.5394\n",
      "Epoch 2/20\n",
      "21675/21675 [==============================] - 801s 37ms/step - loss: 0.1546 - binary_accuracy: 0.7764\n",
      "Epoch 3/20\n",
      "21675/21675 [==============================] - 809s 37ms/step - loss: 0.1286 - binary_accuracy: 0.8216\n",
      "Epoch 4/20\n",
      "21675/21675 [==============================] - 810s 37ms/step - loss: 0.1063 - binary_accuracy: 0.8581\n",
      "Epoch 5/20\n",
      "21675/21675 [==============================] - 801s 37ms/step - loss: 0.0886 - binary_accuracy: 0.8845\n",
      "Epoch 6/20\n",
      "21675/21675 [==============================] - 787s 36ms/step - loss: 0.0765 - binary_accuracy: 0.9009\n",
      "Epoch 7/20\n",
      "21675/21675 [==============================] - 795s 37ms/step - loss: 0.0710 - binary_accuracy: 0.9109\n",
      "Epoch 8/20\n",
      "21675/21675 [==============================] - 792s 37ms/step - loss: 0.0651 - binary_accuracy: 0.9182\n",
      "Epoch 9/20\n",
      "21675/21675 [==============================] - 811s 37ms/step - loss: 0.0622 - binary_accuracy: 0.9222\n",
      "Epoch 10/20\n",
      "21675/21675 [==============================] - 815s 38ms/step - loss: 0.0580 - binary_accuracy: 0.9290\n",
      "Epoch 11/20\n",
      "21675/21675 [==============================] - 823s 38ms/step - loss: 0.0547 - binary_accuracy: 0.9331\n",
      "Epoch 12/20\n",
      "21675/21675 [==============================] - 819s 38ms/step - loss: 0.0508 - binary_accuracy: 0.9383\n",
      "Epoch 13/20\n",
      "21675/21675 [==============================] - 827s 38ms/step - loss: 0.0474 - binary_accuracy: 0.9437\n",
      "Epoch 14/20\n",
      "21675/21675 [==============================] - 826s 38ms/step - loss: 0.0479 - binary_accuracy: 0.9409\n",
      "Epoch 15/20\n",
      "21675/21675 [==============================] - 826s 38ms/step - loss: 0.0490 - binary_accuracy: 0.9410\n",
      "Epoch 16/20\n",
      "21675/21675 [==============================] - 844s 39ms/step - loss: 0.0461 - binary_accuracy: 0.9441\n",
      "Epoch 17/20\n",
      "21675/21675 [==============================] - 833s 38ms/step - loss: 0.0436 - binary_accuracy: 0.9478\n",
      "Epoch 18/20\n",
      "21675/21675 [==============================] - 829s 38ms/step - loss: 0.0438 - binary_accuracy: 0.9472\n",
      "Epoch 19/20\n",
      "21675/21675 [==============================] - 823s 38ms/step - loss: 0.0571 - binary_accuracy: 0.9288\n",
      "Epoch 20/20\n",
      "21675/21675 [==============================] - 819s 38ms/step - loss: 0.0515 - binary_accuracy: 0.9355\n"
     ]
    }
   ],
   "source": [
    "model_batch_hist = model_batch.fit(x_train, np.asarray(y_train), batch_size=n_batch, epochs=n_epoch, verbose=1, shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "AOalT8xDZtW6",
    "outputId": "8be84f8c-20a6-47ab-cd59-e83d38b71a6a"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3825/3825 [==============================] - 33s 9ms/step - loss: 0.1648 - binary_accuracy: 0.7969\n",
      "\n",
      "Model Accuracy: 79.69%\n"
     ]
    }
   ],
   "source": [
    "score_batch = model_batch.evaluate(x_val, np.asarray(y_val), batch_size=n_batch)\n",
    "print(\"\\nModel Accuracy:\", str(round(score_batch[1] * 100, 2)) + '%')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 299
    },
    "id": "lKVFVJUOZtW6",
    "outputId": "6cfb202e-d7a1-4dac-fe21-8e4fb17d2eaf"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.style.use('ggplot')\n",
    "plt.plot(model_batch_hist.history['loss'])\n",
    "\n",
    "plt.title('Model Loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train'], loc='lower left')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "xjS7oADVZtW7"
   },
   "source": [
    "The results are good, let's save this model and try a different approach."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "ELVCgR75ZtW7"
   },
   "outputs": [],
   "source": [
    "model_batch.save('models/LSTM-Batch-Model.h5')  # save model\n",
    "del model_batch  # delete existing model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "id": "Pk-NNsaQaAs7"
   },
   "outputs": [],
   "source": [
    "from keras.models import load_model\n",
    "\n",
    "# returns a compiled model, identical to the previous one\n",
    "model_batch = load_model('models/LSTM-Batch-Model.h5')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "6PzImzbOZtW7"
   },
   "source": [
    "Let's check if everything is correct:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "jkjeuptqZtW7",
    "outputId": "42a0e639-73b4-4808-f244-34ec7688f449"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "30/30 [==============================] - 16s 397ms/step - loss: 0.0692 - binary_accuracy: 0.9182\n",
      "\n",
      "Model Accuracy: 91.82%\n"
     ]
    }
   ],
   "source": [
    "score_batch = model_batch.evaluate(x_val, np.asarray(y_val), batch_size=n_batch)\n",
    "print(\"\\nModel Accuracy:\", str(round(score_batch[1] * 100, 2)) + '%')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "B6IZ4t9mZtW7"
   },
   "source": [
    "<h2 align=\"center\">LSTM Model (Mini-Batch Gradient Descent)</h2>\n",
    "\n",
    "Mini-batch gradient descent is a variation of the gradient descent algorithm that splits the training dataset into small batches that are used to calculate model error and update model coefficients.\n",
    "\n",
    "Mini batch algorithm is the most favorable and widely used algorithm that makes precise and faster results using a batch of `m` training examples. In mini batch algorithm rather than using  the complete data set, in every iteration we use a set of `m` training examples called batch to compute the gradient of the cost function. Common mini-batch sizes range between 32 and 256."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "SPev6XYLZtW8"
   },
   "source": [
    "<h3 align=\"center\">Setting Up and Creating the Network</h3>\n",
    "\n",
    "We will use an LSTM network fit for 20 epochs and with batch size = 128.\n",
    "\n",
    "A mean squared error optimization function is used for this problem with the efficient ADAM optimization algorithm.\n",
    "\n",
    "So now let's configure and create the network:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "a5odeM8hZtW8",
    "outputId": "64298c6b-e27d-4b86-8911-3278ebbbdd03"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " lstm (LSTM)                 (None, 128)               117248    \n",
      "                                                                 \n",
      " dense (Dense)               (None, 1)                 129       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 117,377\n",
      "Trainable params: 117,377\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "# configure network\n",
    "n_neurons = 128\n",
    "n_epoch = 20\n",
    "n_batch = 128\n",
    "\n",
    "# design network\n",
    "model_mini = Sequential()\n",
    "model_mini.add(LSTM(n_neurons, input_shape=(x.shape[1], x.shape[2])))\n",
    "model_mini.add(Dense(1, activation='sigmoid'))\n",
    "model_mini.compile(loss='mean_squared_error', optimizer='adam', metrics=['binary_accuracy'])\n",
    "\n",
    "print(model_mini.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "O1cvbA9gZtW8",
    "outputId": "b7b648a6-9099-4d29-de6c-5d76dedd0402"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "170/170 [==============================] - 60s 338ms/step - loss: 0.2163 - binary_accuracy: 0.6496\n",
      "Epoch 2/20\n",
      "170/170 [==============================] - 57s 337ms/step - loss: 0.1772 - binary_accuracy: 0.7510\n",
      "Epoch 3/20\n",
      "170/170 [==============================] - 56s 330ms/step - loss: 0.1597 - binary_accuracy: 0.7737\n",
      "Epoch 4/20\n",
      "170/170 [==============================] - 56s 330ms/step - loss: 0.1473 - binary_accuracy: 0.7918\n",
      "Epoch 5/20\n",
      "170/170 [==============================] - 56s 332ms/step - loss: 0.1373 - binary_accuracy: 0.8092\n",
      "Epoch 6/20\n",
      "170/170 [==============================] - 57s 338ms/step - loss: 0.1310 - binary_accuracy: 0.8217\n",
      "Epoch 7/20\n",
      "170/170 [==============================] - 57s 335ms/step - loss: 0.1205 - binary_accuracy: 0.8371\n",
      "Epoch 8/20\n",
      "170/170 [==============================] - 57s 333ms/step - loss: 0.1133 - binary_accuracy: 0.8477\n",
      "Epoch 9/20\n",
      "170/170 [==============================] - 57s 336ms/step - loss: 0.1110 - binary_accuracy: 0.8543\n",
      "Epoch 10/20\n",
      "170/170 [==============================] - 57s 334ms/step - loss: 0.1022 - binary_accuracy: 0.8645\n",
      "Epoch 11/20\n",
      "170/170 [==============================] - 57s 337ms/step - loss: 0.0969 - binary_accuracy: 0.8742\n",
      "Epoch 12/20\n",
      "170/170 [==============================] - 58s 342ms/step - loss: 0.0909 - binary_accuracy: 0.8850\n",
      "Epoch 13/20\n",
      "170/170 [==============================] - 62s 362ms/step - loss: 0.0817 - binary_accuracy: 0.8969\n",
      "Epoch 14/20\n",
      "170/170 [==============================] - 58s 342ms/step - loss: 0.0771 - binary_accuracy: 0.9040\n",
      "Epoch 15/20\n",
      "170/170 [==============================] - 57s 334ms/step - loss: 0.0688 - binary_accuracy: 0.9165\n",
      "Epoch 16/20\n",
      "170/170 [==============================] - 59s 345ms/step - loss: 0.0679 - binary_accuracy: 0.9180\n",
      "Epoch 17/20\n",
      "170/170 [==============================] - 60s 352ms/step - loss: 0.0620 - binary_accuracy: 0.9261\n",
      "Epoch 18/20\n",
      "170/170 [==============================] - 57s 332ms/step - loss: 0.0588 - binary_accuracy: 0.9310\n",
      "Epoch 19/20\n",
      "170/170 [==============================] - 57s 336ms/step - loss: 0.0555 - binary_accuracy: 0.9367\n",
      "Epoch 20/20\n",
      "170/170 [==============================] - 57s 337ms/step - loss: 0.0590 - binary_accuracy: 0.9327\n"
     ]
    }
   ],
   "source": [
    "model_mini_hist = model_mini.fit(x_train, np.asarray(y_train), batch_size=n_batch, epochs=n_epoch, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "e4LOkl1DZtW8",
    "outputId": "96ead9b2-a8c5-4e36-c3c1-6de847fcdc3f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "30/30 [==============================] - 4s 127ms/step - loss: 0.1313 - binary_accuracy: 0.8327\n",
      "\n",
      "Model Accuracy: 83.27%\n"
     ]
    }
   ],
   "source": [
    "model_mini_score = model_mini.evaluate(x_val, np.asarray(y_val), batch_size=n_batch)\n",
    "print(\"\\nModel Accuracy:\", str(round(model_mini_score[1] * 100, 2)) + '%')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 295
    },
    "id": "n_q09UyIZtW9",
    "outputId": "ebe863fd-91cd-4da5-9d19-a880844545b7"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(model_mini_hist.history['loss'])\n",
    "\n",
    "plt.title('Model Loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train'], loc='lower left')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "hceEHQ59ZtW9"
   },
   "source": [
    "This is a good value for the model, we can use it. Let's save our model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "RuxZrCz0ZtW-"
   },
   "outputs": [],
   "source": [
    "model_mini.save('models/LSTM-Mini-Batch-Model.h5')  # save model\n",
    "del model_mini  # delete existing model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "id": "9Y-JG9FAZ89D"
   },
   "outputs": [],
   "source": [
    "# returns a compiled model, identical to the previous one\n",
    "model_mini = load_model('models/LSTM-Mini-Batch-Model.h5')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "9FJIKNXBZtW-"
   },
   "source": [
    "Let's check if everything is correct:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "SWZhq9u5ZtW-",
    "outputId": "eb9d4eef-a55b-4cc3-e180-35d7bdd6ca95"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "30/30 [==============================] - 5s 131ms/step - loss: 0.0747 - binary_accuracy: 0.9067\n",
      "\n",
      "Model Accuracy: 90.67%\n"
     ]
    }
   ],
   "source": [
    "model_mini_score = model_mini.evaluate(x_val, np.asarray(y_val), batch_size=n_batch)\n",
    "print(\"\\nModel Accuracy:\", str(round(model_mini_score[1] * 100, 2)) + '%')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Hhi8zre1ZtW-"
   },
   "source": [
    "<h2 align=\"center\">Stacked LSTM Layers</h2>\n",
    "\n",
    "Now we will try the model with a 3 stacked LSTM layers:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "-rhE_r9JZtW_",
    "outputId": "70b98c21-7ef9-4616-dd71-96e827fb661f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " lstm (LSTM)                 (None, 100, 100)          80400     \n",
      "                                                                 \n",
      " lstm_1 (LSTM)               (None, 100, 32)           17024     \n",
      "                                                                 \n",
      " lstm_2 (LSTM)               (None, 32)                8320      \n",
      "                                                                 \n",
      " dense (Dense)               (None, 1)                 33        \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 105,777\n",
      "Trainable params: 105,777\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "# configure network\n",
    "n_neurons = 100\n",
    "n_epoch = 20\n",
    "n_batch = 128\n",
    "\n",
    "# design network\n",
    "model3L = Sequential()\n",
    "\n",
    "model3L.add(LSTM(n_neurons, return_sequences=True, input_shape=(x.shape[1], x.shape[2])))\n",
    "model3L.add(LSTM(32, return_sequences=True))  # returns a sequence of vectors of dimension 32\n",
    "model3L.add(LSTM(32))  # return a single vector of dimension 32\n",
    "\n",
    "model3L.add(Dense(1, activation='sigmoid'))\n",
    "model3L.compile(loss='mean_squared_error', optimizer='adam', metrics=['binary_accuracy'])\n",
    "\n",
    "print(model3L.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "rfk9cxSYZtW_",
    "outputId": "6b2061e9-7c0f-406a-c45b-cd4df26a01f8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "170/170 [==============================] - 74s 399ms/step - loss: 0.2189 - binary_accuracy: 0.6408\n",
      "Epoch 2/20\n",
      "170/170 [==============================] - 69s 408ms/step - loss: 0.1775 - binary_accuracy: 0.7501\n",
      "Epoch 3/20\n",
      "170/170 [==============================] - 68s 398ms/step - loss: 0.1627 - binary_accuracy: 0.7672\n",
      "Epoch 4/20\n",
      "170/170 [==============================] - 73s 431ms/step - loss: 0.1502 - binary_accuracy: 0.7873\n",
      "Epoch 5/20\n",
      "170/170 [==============================] - 68s 403ms/step - loss: 0.1473 - binary_accuracy: 0.7971\n",
      "Epoch 6/20\n",
      "170/170 [==============================] - 70s 411ms/step - loss: 0.1373 - binary_accuracy: 0.8132\n",
      "Epoch 7/20\n",
      "170/170 [==============================] - 68s 403ms/step - loss: 0.1293 - binary_accuracy: 0.8238\n",
      "Epoch 8/20\n",
      "170/170 [==============================] - 68s 398ms/step - loss: 0.1225 - binary_accuracy: 0.8332\n",
      "Epoch 9/20\n",
      "170/170 [==============================] - 68s 401ms/step - loss: 0.1131 - binary_accuracy: 0.8470\n",
      "Epoch 10/20\n",
      "170/170 [==============================] - 69s 407ms/step - loss: 0.1132 - binary_accuracy: 0.8470\n",
      "Epoch 11/20\n",
      "170/170 [==============================] - 70s 410ms/step - loss: 0.1026 - binary_accuracy: 0.8633\n",
      "Epoch 12/20\n",
      "170/170 [==============================] - 68s 402ms/step - loss: 0.0968 - binary_accuracy: 0.8743\n",
      "Epoch 13/20\n",
      "170/170 [==============================] - 69s 404ms/step - loss: 0.0919 - binary_accuracy: 0.8824\n",
      "Epoch 14/20\n",
      "170/170 [==============================] - 69s 406ms/step - loss: 0.0840 - binary_accuracy: 0.8930\n",
      "Epoch 15/20\n",
      "170/170 [==============================] - 69s 408ms/step - loss: 0.0828 - binary_accuracy: 0.8957\n",
      "Epoch 16/20\n",
      "170/170 [==============================] - 69s 407ms/step - loss: 0.0732 - binary_accuracy: 0.9098\n",
      "Epoch 17/20\n",
      "170/170 [==============================] - 69s 404ms/step - loss: 0.0701 - binary_accuracy: 0.9145\n",
      "Epoch 18/20\n",
      "170/170 [==============================] - 68s 403ms/step - loss: 0.0645 - binary_accuracy: 0.9230\n",
      "Epoch 19/20\n",
      "170/170 [==============================] - 69s 407ms/step - loss: 0.0626 - binary_accuracy: 0.9246\n",
      "Epoch 20/20\n",
      "170/170 [==============================] - 69s 407ms/step - loss: 0.0625 - binary_accuracy: 0.9258\n"
     ]
    }
   ],
   "source": [
    "model3L_hist = model3L.fit(x_train, np.asarray(y_train), batch_size=n_batch, epochs=n_epoch, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "Z66lHsYnZtW_",
    "outputId": "d727be3d-2058-4266-df4e-c55d2b6bd3cd"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "30/30 [==============================] - 6s 147ms/step - loss: 0.1419 - binary_accuracy: 0.8207\n",
      "\n",
      "Model Accuracy: 82.07%\n"
     ]
    }
   ],
   "source": [
    "model3L_score = model3L.evaluate(x_val, np.asarray(y_val), batch_size=n_batch)\n",
    "print(\"\\nModel Accuracy:\", str(round(model3L_score[1] * 100, 2)) + '%')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 295
    },
    "id": "tx4LX6JNZtW_",
    "outputId": "ddf63d0b-15e2-491a-a274-2df5e6c17648"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(model3L_hist.history['loss'])\n",
    "\n",
    "plt.title('Model Loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train'], loc='lower left')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "vTuyvAM-ZtXC"
   },
   "source": [
    "Let's save our model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "QCG_Af_EZtXC"
   },
   "outputs": [],
   "source": [
    "model3L.save('models/LSTM-3L-Model.h5')  # save model\n",
    "del model3L  # delete existing model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "id": "Tgs_iFa1aKdP"
   },
   "outputs": [],
   "source": [
    "# returns a compiled model, identical to the previous one\n",
    "model3L = load_model('models/LSTM-3L-Model.h5')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "l4ATCsPeZtXC"
   },
   "source": [
    "Let's check if everything is correct:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "OA83RjgYZtXC",
    "outputId": "473f5f6f-f1bf-4039-91a3-87d801148a48"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "30/30 [==============================] - 6s 144ms/step - loss: 0.0740 - binary_accuracy: 0.9098\n",
      "\n",
      "Model Accuracy: 90.98%\n"
     ]
    }
   ],
   "source": [
    "model3L_score = model3L.evaluate(x_val, np.asarray(y_val), batch_size=n_batch)\n",
    "print(\"\\nModel Accuracy:\", str(round(model3L_score[1] * 100, 2)) + '%')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "d2zteJmhZtXC"
   },
   "source": [
    "<h2 align=\"center\">Bidirectional LSTM</h2>\n",
    "\n",
    "Bidirectional LSTMs are an extension of traditional LSTMs that can improve model performance on sequence classification problems.\n",
    "\n",
    "In problems where all timesteps of the input sequence are available, Bidirectional LSTMs train two instead of one LSTMs on the input sequence. The first on the input sequence as-is and the second on a reversed copy of the input sequence. This can provide additional context to the network and result in faster and even fuller learning on the problem.\n",
    "\n",
    "The idea of Bidirectional Recurrent Neural Networks (RNNs) is straightforward.\n",
    "\n",
    "It involves duplicating the first recurrent layer in the network so that there are now two layers side-by-side, then providing the input sequence as-is as input to the first layer and providing a reversed copy of the input sequence to the second."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "91_ZiDYhZtXD",
    "outputId": "315ee5f3-8ff7-4ced-e6cf-d6c5d0710c40"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " bidirectional (Bidirectiona  (None, 100, 128)         84480     \n",
      " l)                                                              \n",
      "                                                                 \n",
      " bidirectional_1 (Bidirectio  (None, 128)              98816     \n",
      " nal)                                                            \n",
      "                                                                 \n",
      " dropout (Dropout)           (None, 128)               0         \n",
      "                                                                 \n",
      " dense_1 (Dense)             (None, 1)                 129       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 183,425\n",
      "Trainable params: 183,425\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "from keras.layers.core import Dropout\n",
    "from keras.layers import Bidirectional\n",
    "\n",
    "# configure network\n",
    "n_epoch = 20\n",
    "n_batch = 64\n",
    "\n",
    "# design network\n",
    "model_Bi_LSTM = Sequential()\n",
    "model_Bi_LSTM.add(Bidirectional(LSTM(64, return_sequences=True), input_shape=(x.shape[1], x.shape[2])))\n",
    "model_Bi_LSTM.add(Bidirectional(LSTM(64)))\n",
    "model_Bi_LSTM.add(Dropout(0.5))\n",
    "model_Bi_LSTM.add(Dense(1, activation='sigmoid'))\n",
    "\n",
    "# try using different optimizers and different optimizer configs\n",
    "model_Bi_LSTM.compile('adam', 'binary_crossentropy', metrics=['binary_accuracy'])\n",
    "print(model_Bi_LSTM.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "FUBS8nSHZtXD",
    "outputId": "830d52e7-b26c-40d2-f15c-2aa5c55f1f8a"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "339/339 [==============================] - 133s 370ms/step - loss: 0.5575 - binary_accuracy: 0.7166 - val_loss: 0.5054 - val_binary_accuracy: 0.7537\n",
      "Epoch 2/20\n",
      "339/339 [==============================] - 125s 369ms/step - loss: 0.4757 - binary_accuracy: 0.7772 - val_loss: 0.4474 - val_binary_accuracy: 0.7856\n",
      "Epoch 3/20\n",
      "339/339 [==============================] - 123s 363ms/step - loss: 0.4388 - binary_accuracy: 0.7957 - val_loss: 0.4401 - val_binary_accuracy: 0.7864\n",
      "Epoch 4/20\n",
      "339/339 [==============================] - 121s 357ms/step - loss: 0.4008 - binary_accuracy: 0.8215 - val_loss: 0.4184 - val_binary_accuracy: 0.8026\n",
      "Epoch 5/20\n",
      "339/339 [==============================] - 121s 356ms/step - loss: 0.3726 - binary_accuracy: 0.8356 - val_loss: 0.3843 - val_binary_accuracy: 0.8303\n",
      "Epoch 6/20\n",
      "339/339 [==============================] - 120s 355ms/step - loss: 0.3490 - binary_accuracy: 0.8478 - val_loss: 0.3800 - val_binary_accuracy: 0.8340\n",
      "Epoch 7/20\n",
      "339/339 [==============================] - 121s 357ms/step - loss: 0.3240 - binary_accuracy: 0.8627 - val_loss: 0.4203 - val_binary_accuracy: 0.8081\n",
      "Epoch 8/20\n",
      "339/339 [==============================] - 122s 359ms/step - loss: 0.3008 - binary_accuracy: 0.8763 - val_loss: 0.3766 - val_binary_accuracy: 0.8358\n",
      "Epoch 9/20\n",
      "339/339 [==============================] - 125s 370ms/step - loss: 0.2718 - binary_accuracy: 0.8914 - val_loss: 0.4048 - val_binary_accuracy: 0.8327\n",
      "Epoch 10/20\n",
      "339/339 [==============================] - 125s 370ms/step - loss: 0.2536 - binary_accuracy: 0.8975 - val_loss: 0.4587 - val_binary_accuracy: 0.8254\n",
      "Epoch 11/20\n",
      "339/339 [==============================] - 123s 362ms/step - loss: 0.2269 - binary_accuracy: 0.9116 - val_loss: 0.4401 - val_binary_accuracy: 0.8345\n",
      "Epoch 12/20\n",
      "339/339 [==============================] - 123s 362ms/step - loss: 0.1947 - binary_accuracy: 0.9257 - val_loss: 0.4250 - val_binary_accuracy: 0.8439\n",
      "Epoch 13/20\n",
      "339/339 [==============================] - 123s 362ms/step - loss: 0.1760 - binary_accuracy: 0.9324 - val_loss: 0.4879 - val_binary_accuracy: 0.8395\n",
      "Epoch 14/20\n",
      "339/339 [==============================] - 122s 359ms/step - loss: 0.1466 - binary_accuracy: 0.9453 - val_loss: 0.5243 - val_binary_accuracy: 0.8371\n",
      "Epoch 15/20\n",
      "339/339 [==============================] - 123s 364ms/step - loss: 0.1220 - binary_accuracy: 0.9562 - val_loss: 0.5274 - val_binary_accuracy: 0.8217\n",
      "Epoch 16/20\n",
      "339/339 [==============================] - 123s 364ms/step - loss: 0.1220 - binary_accuracy: 0.9564 - val_loss: 0.6252 - val_binary_accuracy: 0.8259\n",
      "Epoch 17/20\n",
      "339/339 [==============================] - 124s 365ms/step - loss: 0.0949 - binary_accuracy: 0.9682 - val_loss: 0.5920 - val_binary_accuracy: 0.8337\n",
      "Epoch 18/20\n",
      "339/339 [==============================] - 121s 358ms/step - loss: 0.0744 - binary_accuracy: 0.9757 - val_loss: 0.7071 - val_binary_accuracy: 0.8188\n",
      "Epoch 19/20\n",
      "339/339 [==============================] - 121s 358ms/step - loss: 0.0818 - binary_accuracy: 0.9727 - val_loss: 0.6927 - val_binary_accuracy: 0.8235\n",
      "Epoch 20/20\n",
      "339/339 [==============================] - 122s 361ms/step - loss: 0.0703 - binary_accuracy: 0.9761 - val_loss: 0.7010 - val_binary_accuracy: 0.8342\n"
     ]
    }
   ],
   "source": [
    "model_Bi_LSTM_hist = model_Bi_LSTM.fit(x_train, np.asarray(y_train), epochs=n_epoch, batch_size=n_batch, verbose=1, validation_data=(x_val, np.asarray(y_val)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "S46I1EENZtXD",
    "outputId": "24c2e9ca-71e0-478a-873e-f002857f0fad"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "60/60 [==============================] - 6s 105ms/step - loss: 0.7010 - binary_accuracy: 0.8342\n",
      "\n",
      "Model Accuracy: 83.42%\n"
     ]
    }
   ],
   "source": [
    "model_Bi_LSTM_score = model_Bi_LSTM.evaluate(x_val, np.asarray(y_val), batch_size=n_batch)\n",
    "print(\"\\nModel Accuracy:\",str(round(model_Bi_LSTM_score[1] * 100, 2)) + '%')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 295
    },
    "id": "IVyY3L1tZtXD",
    "outputId": "23484dd3-e718-475a-e6bf-792b2d47f72c"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(model_Bi_LSTM_hist.history['loss'])\n",
    "plt.plot(model_Bi_LSTM_hist.history['val_loss'])\n",
    "\n",
    "plt.title('Model Loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Val'], loc='lower left')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 295
    },
    "id": "rZOeE_vLZtXD",
    "outputId": "ec579766-0b8d-4999-c8cb-684c019ddf13"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(model_Bi_LSTM_hist.history['binary_accuracy'])\n",
    "plt.plot(model_Bi_LSTM_hist.history['val_binary_accuracy'])\n",
    "\n",
    "plt.title('Model Loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Val'], loc='lower left')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "IoQJQeIdZtXE"
   },
   "source": [
    "Let's save our model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "kscMkhQkZtXE"
   },
   "outputs": [],
   "source": [
    "model_Bi_LSTM.save('models/LSTM-Bi-Model.h5')  # save model\n",
    "del model_Bi_LSTM  # delete existing model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "id": "d0YonJWraOeF"
   },
   "outputs": [],
   "source": [
    "# returns a compiled model, identical to the previous one\n",
    "model_Bi_LSTM = load_model('models/LSTM-Bi-Model.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "7TDowu1umwzy",
    "outputId": "2cc9ebf8-8665-4788-e094-b4d0b332e930"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "60/60 [==============================] - 7s 93ms/step - loss: 0.2522 - binary_accuracy: 0.9331\n",
      "\n",
      "Model Accuracy: 93.31%\n"
     ]
    }
   ],
   "source": [
    "model_Bi_LSTM_score = model_Bi_LSTM.evaluate(x_val, np.asarray(y_val), batch_size=n_batch)\n",
    "print(\"\\nModel Accuracy:\", str(round(model_Bi_LSTM_score[1] * 100, 2)) + '%')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "a3vslMkOZtXE"
   },
   "source": [
    "<h2 align=\"center\">CNN + LSTM</h2>\n",
    "\n",
    "This approach is based entirely on [this work](https://www.aclweb.org/anthology/P16-2037.pdf). As indicated in the work:\n",
    "\n",
    "<cite>\n",
    "    \n",
    "    Dimensional sentiment analysis aims to recognize continuous numerical values in multiple dimensions such as the valencearousal (VA) space. \n",
    "    \n",
    "    Compared to the categorical approach that focuses on sentiment classification such as binary classification (i.e., positive and negative), the dimensional approach can provide more fine-grained sentiment analysis. This study proposes a regional CNN-LSTM model consisting of two parts: regional CNN and LSTM to predict the VA ratings of texts. \n",
    "   \n",
    "    Unlike a conventional CNN which considers a whole text as input, the proposed regional CNN uses an individual sentence as a region, dividing an input text into several regions such that the useful affective information in each region can be extracted and weighted according to their contribution to the VA prediction. Such regional information is sequentially integrated across regions using LSTM for VA prediction. \n",
    "\n",
    "</cite>\n",
    "\n",
    "In short, word vectors of vocabulary words are trained from a large corpus using the word2vec toolkit. For each given text, the  regional CNN model uses the sentence as a region to divide the text into R-domains, ie r1, ..., ri rj, rk, ..., rR. In each region, useful affective functions can be removed when word vectors pass sequentially through the convolutional layer and the  maxpooling  layer. Then such local (regional) features are sequentially integrated between regions using LSTM to construct a text vector for VA prediction."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "DwNuefZ_ZtXE",
    "outputId": "ae34fb87-838b-49a1-f2cd-d470acf04831"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " conv1d (Conv1D)             (None, 100, 128)          51328     \n",
      "                                                                 \n",
      " max_pooling1d (MaxPooling1D  (None, 33, 128)          0         \n",
      " )                                                               \n",
      "                                                                 \n",
      " bidirectional_2 (Bidirectio  (None, 33, 128)          98816     \n",
      " nal)                                                            \n",
      "                                                                 \n",
      " lstm_6 (LSTM)               (None, 32)                20608     \n",
      "                                                                 \n",
      " dropout_1 (Dropout)         (None, 32)                0         \n",
      "                                                                 \n",
      " dense_2 (Dense)             (None, 1)                 33        \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 170,785\n",
      "Trainable params: 170,785\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "from keras.layers import Conv1D, MaxPooling1D, Flatten\n",
    "from keras.layers import Bidirectional\n",
    "\n",
    "# configure network\n",
    "n_epoch = 20\n",
    "n_batch = 128\n",
    "\n",
    "# design network\n",
    "model_CNN_LSTM = Sequential()\n",
    "\n",
    "model_CNN_LSTM.add(Conv1D(filters=128, kernel_size=4, input_shape=(x.shape[1], x.shape[2]), padding='same', activation='relu'))\n",
    "model_CNN_LSTM.add(MaxPooling1D(pool_size=3))\n",
    "model_CNN_LSTM.add(Bidirectional(LSTM(64, return_sequences = True)))\n",
    "model_CNN_LSTM.add(LSTM(32, recurrent_dropout = 0.4))\n",
    "model_CNN_LSTM.add(Dropout(0.2))\n",
    "model_CNN_LSTM.add(Dense(1, activation='sigmoid'))\n",
    "\n",
    "model_CNN_LSTM.compile(loss='binary_crossentropy', optimizer='adam', metrics=['binary_accuracy'])\n",
    "\n",
    "print(model_CNN_LSTM.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "nG3Gl2KlZtXE",
    "outputId": "0b161fe6-0dbb-4c5a-ed67-a4195d090381"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "170/170 [==============================] - 50s 252ms/step - loss: 0.5844 - binary_accuracy: 0.6836\n",
      "Epoch 2/20\n",
      "170/170 [==============================] - 43s 254ms/step - loss: 0.4661 - binary_accuracy: 0.7814\n",
      "Epoch 3/20\n",
      "170/170 [==============================] - 43s 254ms/step - loss: 0.3921 - binary_accuracy: 0.8293\n",
      "Epoch 4/20\n",
      "170/170 [==============================] - 43s 253ms/step - loss: 0.3271 - binary_accuracy: 0.8630\n",
      "Epoch 5/20\n",
      "170/170 [==============================] - 43s 252ms/step - loss: 0.2834 - binary_accuracy: 0.8844\n",
      "Epoch 6/20\n",
      "170/170 [==============================] - 43s 253ms/step - loss: 0.2306 - binary_accuracy: 0.9112\n",
      "Epoch 7/20\n",
      "170/170 [==============================] - 44s 257ms/step - loss: 0.1780 - binary_accuracy: 0.9364\n",
      "Epoch 8/20\n",
      "170/170 [==============================] - 43s 255ms/step - loss: 0.1597 - binary_accuracy: 0.9436\n",
      "Epoch 9/20\n",
      "170/170 [==============================] - 44s 256ms/step - loss: 0.1218 - binary_accuracy: 0.9592\n",
      "Epoch 10/20\n",
      "170/170 [==============================] - 44s 261ms/step - loss: 0.1073 - binary_accuracy: 0.9650\n",
      "Epoch 11/20\n",
      "170/170 [==============================] - 44s 260ms/step - loss: 0.0918 - binary_accuracy: 0.9720\n",
      "Epoch 12/20\n",
      "170/170 [==============================] - 44s 259ms/step - loss: 0.0876 - binary_accuracy: 0.9728\n",
      "Epoch 13/20\n",
      "170/170 [==============================] - 44s 258ms/step - loss: 0.0729 - binary_accuracy: 0.9791\n",
      "Epoch 14/20\n",
      "170/170 [==============================] - 44s 258ms/step - loss: 0.0639 - binary_accuracy: 0.9821\n",
      "Epoch 15/20\n",
      "170/170 [==============================] - 44s 259ms/step - loss: 0.0612 - binary_accuracy: 0.9844\n",
      "Epoch 16/20\n",
      "170/170 [==============================] - 45s 263ms/step - loss: 0.0551 - binary_accuracy: 0.9854\n",
      "Epoch 17/20\n",
      "170/170 [==============================] - 44s 259ms/step - loss: 0.0544 - binary_accuracy: 0.9855\n",
      "Epoch 18/20\n",
      "170/170 [==============================] - 44s 260ms/step - loss: 0.0503 - binary_accuracy: 0.9871\n",
      "Epoch 19/20\n",
      "170/170 [==============================] - 44s 258ms/step - loss: 0.0934 - binary_accuracy: 0.9687\n",
      "Epoch 20/20\n",
      "170/170 [==============================] - 44s 258ms/step - loss: 0.0445 - binary_accuracy: 0.9890\n"
     ]
    }
   ],
   "source": [
    "model_CNN_LSTM_hist = model_CNN_LSTM.fit(x_train, np.asarray(y_train), epochs=n_epoch, batch_size=n_batch)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "yK8I8-i1ZtXF",
    "outputId": "d8c52db7-4e20-45a8-9753-800407b10020"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "30/30 [==============================] - 4s 81ms/step - loss: 0.8386 - binary_accuracy: 0.8188\n",
      "\n",
      "Model Accuracy: 81.88%\n"
     ]
    }
   ],
   "source": [
    "model_CNN_LSTM_score = model_CNN_LSTM.evaluate(x_val, np.asarray(y_val), batch_size=n_batch)\n",
    "print(\"\\nModel Accuracy:\",str(round(model_CNN_LSTM_score[1] * 100, 2)) + '%')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 295
    },
    "id": "Zyk3Q0tXZtXF",
    "outputId": "103996cc-c44c-4b63-dca6-fd439090b282"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(model_CNN_LSTM_hist.history['loss'])\n",
    "\n",
    "plt.title('Model Loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train'], loc='lower left')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "mbH-cK4TZtXF"
   },
   "source": [
    "Let's save our model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "nXJekFFUZtXF"
   },
   "outputs": [],
   "source": [
    "model_CNN_LSTM.save('models/LSTM-CNN-Model.h5')  # save model\n",
    "del model_CNN_LSTM  # delete existing model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "id": "P1p_mEdwpEkK"
   },
   "outputs": [],
   "source": [
    "# returns a compiled model, identical to the previous one\n",
    "model_CNN_LSTM = load_model('models/LSTM-CNN-Model.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "uhvSNNC5m9ky",
    "outputId": "f82cd583-1870-491d-e9fa-14e2c8c795df"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "30/30 [==============================] - 5s 75ms/step - loss: 0.2533 - binary_accuracy: 0.9448\n",
      "\n",
      "Model Accuracy: 94.48%\n"
     ]
    }
   ],
   "source": [
    "model_CNN_LSTM_score = model_CNN_LSTM.evaluate(x_val, np.asarray(y_val), batch_size=n_batch)\n",
    "print(\"\\nModel Accuracy:\", str(round(model_CNN_LSTM_score[1] * 100, 2)) + '%')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "shEEexHiZtXF"
   },
   "source": [
    "<h2 align=\"center\">Testing</h2>\n",
    "\n",
    "So far in terms of accuracy the best model is LSTM Model with Mini-Batch Gradient Descent.\n",
    "\n",
    "Let's try to check the test data with all but one of the test examples to see if the model can predict it:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "6E0ZIkM0ZtXF",
    "outputId": "cd2567c0-0d8e-4897-86a3-a2708272e847"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Length of test data (except 1): 5781\n"
     ]
    }
   ],
   "source": [
    "print(\"Length of test data (except 1):\", len(x_test[:-1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "U3otd9AgZtXG",
    "outputId": "b626800f-716e-439a-af30-5cbc8601f974"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "181/181 [==============================] - 7s 35ms/step - loss: 0.1390 - binary_accuracy: 0.8324\n",
      "[0.13903364539146423, 0.8323819637298584]\n"
     ]
    }
   ],
   "source": [
    "score_test = model_mini.evaluate(x_test[:-1], np.asarray(y_test[:-1]))\n",
    "print(score_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "G5MRX62sZtXG",
    "outputId": "42052f63-6fa6-4ccf-bcda-ea88fecaa766"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1/1 [==============================] - 1s 583ms/step - loss: 0.0084 - binary_accuracy: 1.0000\n",
      "[0.008366141468286514, 1.0]\n"
     ]
    }
   ],
   "source": [
    "score_test_one = model_CNN_LSTM.evaluate(x_test[-1:], np.asarray(y_test[-1:]))\n",
    "print(score_test_one)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "xE4SWgv4ZtXG",
    "outputId": "6e1498cc-cbea-48e9-f58e-4a2901f8db55"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[1]\n",
      " [0]\n",
      " [1]\n",
      " ...\n",
      " [1]\n",
      " [1]\n",
      " [1]]\n"
     ]
    }
   ],
   "source": [
    "predict_test = (model_CNN_LSTM.predict(x_test[:-1]) > 0.5).astype(\"int32\")\n",
    "print(predict_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "HUnlwzSQZtXG"
   },
   "outputs": [],
   "source": [
    "predict_test_one = (model_CNN_LSTM.predict(x_test[-1:]) > 0.5).astype(\"int32\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "y_OgHg7sZtXG",
    "outputId": "8e7eb65c-a0aa-43e5-c145-1fe432d782b9"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Predicted value: 1 \n",
      "Actual value: 1\n"
     ]
    }
   ],
   "source": [
    "print(\"Predicted value:\", predict_test_one[0][0], \"\\nActual value:\", y_test[-1:][0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "2qtrGThgZtXG"
   },
   "source": [
    "Prediction works great, now let's visualize the data."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ohO5DMXQZtXH"
   },
   "source": [
    "<h2 align=\"center\">Visualization of Classification Report</h2>\n",
    "\n",
    "We will need our method from previous part to visualize our data, so we will use them now."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "id": "dwvIw5LMZtXH"
   },
   "outputs": [],
   "source": [
    "def plot_classification_report(classification_report, title='Classification Report', cmap='RdBu'):\n",
    "    \n",
    "    lines = classification_report.split('\\n')\n",
    "\n",
    "    classes, plotMat, support, class_names = [], [], [], []\n",
    "    \n",
    "    for line in lines[2 : (len(lines) - 5)]:\n",
    "        t = line.strip().split()\n",
    "        if len(t) < 2: continue\n",
    "        classes.append(t[0])\n",
    "        v = [float(x) for x in t[1: len(t) - 1]]\n",
    "        support.append(int(t[-1]))\n",
    "        class_names.append(t[0])\n",
    "        plotMat.append(v)\n",
    "\n",
    "    xlabel = 'Metrics'\n",
    "    ylabel = 'Classes'\n",
    "    \n",
    "    xticklabels = ['Precision', 'Recall', 'F1-score']\n",
    "    yticklabels = ['{0} ({1})'.format(class_names[idx], sup) for idx, sup  in enumerate(support)]\n",
    "    \n",
    "    figure_width = 25\n",
    "    figure_height = len(class_names) + 7\n",
    "    correct_orientation = False\n",
    "    \n",
    "    heatmap(np.array(plotMat), title, xlabel, ylabel, xticklabels, yticklabels, figure_width, figure_height, correct_orientation, cmap=cmap)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "id": "-SxQWnH2ZtXH"
   },
   "outputs": [],
   "source": [
    "def heatmap(AUC, title, xlabel, ylabel, xticklabels, yticklabels, figure_width=40, figure_height=20, correct_orientation=False, cmap='RdBu'):\n",
    "    fig, ax = plt.subplots()\n",
    "    c = ax.pcolor(AUC, edgecolors='k', linestyle='dashed', linewidths=0.2, cmap=cmap)\n",
    "\n",
    "    # put the major ticks at the middle of each cell\n",
    "    ax.set_yticks(np.arange(AUC.shape[0]) + 0.5, minor=False)\n",
    "    ax.set_xticks(np.arange(AUC.shape[1]) + 0.5, minor=False)\n",
    "\n",
    "    # set tick labels\n",
    "    ax.set_xticklabels(xticklabels, minor=False)\n",
    "    ax.set_yticklabels(yticklabels, minor=False)\n",
    "\n",
    "    # set title and x/y labels\n",
    "    plt.title(title)\n",
    "    plt.xlabel(xlabel)\n",
    "    plt.ylabel(ylabel)      \n",
    "\n",
    "    # remove last blank column\n",
    "    plt.xlim( (0, AUC.shape[1]) )\n",
    "\n",
    "    # turn off all the ticks\n",
    "    ax = plt.gca()    \n",
    "    for t in ax.xaxis.get_major_ticks():\n",
    "        t.tick1line.set_visible = False\n",
    "        t.tick2line.set_visible = False\n",
    "    for t in ax.yaxis.get_major_ticks():\n",
    "        t.tick1line.set_visible = False\n",
    "        t.tick2line.set_visible = False\n",
    "\n",
    "    # add color bar\n",
    "    plt.colorbar(c)\n",
    "\n",
    "    # add text in each cell \n",
    "    show_val(c)\n",
    "\n",
    "    # proper orientation (origin at the top left instead of bottom left)\n",
    "    if correct_orientation:\n",
    "        ax.invert_yaxis()\n",
    "        ax.xaxis.tick_top()       \n",
    "\n",
    "    # resize \n",
    "    fig = plt.gcf()\n",
    "    fig.set_size_inches(cm_to_inch(figure_width, figure_height))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "id": "jqu5LdXZZtXH"
   },
   "outputs": [],
   "source": [
    "def show_val(pc, fmt=\"%.2f\", **kw):\n",
    "    pc.update_scalarmappable()\n",
    "    ax = pc.axes\n",
    "    for p, color, value in zip(pc.get_paths(), pc.get_facecolors(), pc.get_array()):\n",
    "        x, y = p.vertices[:-2, :].mean(0)\n",
    "        if np.all(color[:3] > 0.5):\n",
    "            color = (0.0, 0.0, 0.0)\n",
    "        else:\n",
    "            color = (1.0, 1.0, 1.0)\n",
    "        ax.text(x, y, fmt % value, ha=\"center\", va=\"center\", color=color, **kw)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "id": "L_QQe0fPZtXH"
   },
   "outputs": [],
   "source": [
    "def cm_to_inch(*dim):\n",
    "    inch = 2.54\n",
    "    return tuple(i/inch for i in dim[0]) if type(dim[0]) == tuple else tuple(i/inch for i in dim)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Fb4DlXofZtXI"
   },
   "source": [
    "<h2 align=\"center\">Comparison of Models</h2> \n",
    "\n",
    "A useful tool when predicting the probability of a binary outcome is the Receiver Operating Characteristic curve, or ROC curve.\n",
    "\n",
    "It is a plot of the false positive rate (x-axis) versus the true positive rate (y-axis) for a number of different candidate threshold values between 0.0 and 1.0. Put another way, it plots the false alarm rate versus the hit rate.\n",
    "\n",
    "The true positive rate is calculated as the number of true positives divided by the sum of the number of true positives and the number of false negatives. It describes how good the model is at predicting the positive class when the actual outcome is positive.\n",
    "\n",
    "The false positive rate is calculated as the number of false positives divided by the sum of the number of false positives and the number of true negatives.\n",
    "\n",
    "It is also called the false alarm rate as it summarizes how often a positive class is predicted when the actual outcome is negative.\n",
    "\n",
    "To make this clear:\n",
    "* Smaller values on the x-axis of the plot indicate lower false positives and higher true negatives.\n",
    "* Larger values on the y-axis of the plot indicate higher true positives and lower false negatives."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 445
    },
    "id": "tbcuMZNGZtXI",
    "outputId": "12db50f3-569e-45e5-d25e-5851c4b8c834"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report LSTM Stochastic Gradient Descent Model:\n",
      "               precision    recall  f1-score   support\n",
      "\n",
      "           0       0.90      0.91      0.91      2176\n",
      "           1       0.92      0.90      0.91      2324\n",
      "\n",
      "    accuracy                           0.91      4500\n",
      "   macro avg       0.91      0.91      0.91      4500\n",
      "weighted avg       0.91      0.91      0.91      4500\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 708.661x255.118 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report\n",
    "\n",
    "batch_prediction = (model_batch.predict(x_test) > 0.5).astype(\"int32\")\n",
    "print('Classification Report LSTM Stochastic Gradient Descent Model:\\n', classification_report(y_test, batch_prediction))\n",
    "\n",
    "plot_classification_report(classification_report(y_test, batch_prediction), title='LSTM Stochastic Gradient Descent Classification Report', cmap='RdBu')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 445
    },
    "id": "o7ZSh7hjZtXI",
    "outputId": "178068e0-72f4-4097-bab6-8f0caa00b672"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report LSTM Mini-Batch Model:\n",
      "               precision    recall  f1-score   support\n",
      "\n",
      "           0       0.90      0.94      0.92      2176\n",
      "           1       0.94      0.90      0.92      2324\n",
      "\n",
      "    accuracy                           0.92      4500\n",
      "   macro avg       0.92      0.92      0.92      4500\n",
      "weighted avg       0.92      0.92      0.92      4500\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 708.661x255.118 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "mini_prediction = (model_mini.predict(x_test) > 0.5).astype(\"int32\")\n",
    "print('Classification Report LSTM Mini-Batch Model:\\n', classification_report(y_test, mini_prediction))\n",
    "\n",
    "plot_classification_report(classification_report(y_test, mini_prediction), title='LSTM Mini-Batch Classification Report', cmap='RdBu')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 445
    },
    "id": "LMHvPMiIZtXI",
    "outputId": "a2c2cd27-fe8c-4f1c-d434-543467c3c4f2"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report LSTM Stacked Layers Model:\n",
      "               precision    recall  f1-score   support\n",
      "\n",
      "           0       0.91      0.89      0.90      2176\n",
      "           1       0.90      0.92      0.91      2324\n",
      "\n",
      "    accuracy                           0.91      4500\n",
      "   macro avg       0.91      0.91      0.91      4500\n",
      "weighted avg       0.91      0.91      0.91      4500\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 708.661x255.118 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "stack_layer_prediction = (model3L.predict(x_test) > 0.5).astype(\"int32\")\n",
    "print('Classification Report LSTM Stacked Layers Model:\\n', classification_report(y_test, stack_layer_prediction))\n",
    "\n",
    "plot_classification_report(classification_report(y_test, stack_layer_prediction), title='LSTM Stacked Layers Classification Report', cmap='RdBu')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 445
    },
    "id": "QW-amKnwZtXI",
    "outputId": "d72b1631-ff1a-4d7c-fb51-8f59aa096d35"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report Bidirectional LSTM Model:\n",
      "               precision    recall  f1-score   support\n",
      "\n",
      "           0       0.95      0.92      0.93      2176\n",
      "           1       0.92      0.96      0.94      2324\n",
      "\n",
      "    accuracy                           0.94      4500\n",
      "   macro avg       0.94      0.94      0.94      4500\n",
      "weighted avg       0.94      0.94      0.94      4500\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 708.661x255.118 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "lstm_bi_prediction = (model_Bi_LSTM.predict(x_test) > 0.5).astype(\"int32\")\n",
    "print('Classification Report Bidirectional LSTM Model:\\n', classification_report(y_test, lstm_bi_prediction))\n",
    "\n",
    "plot_classification_report(classification_report(y_test, lstm_bi_prediction), title='Bidirectional LSTM Classification Report', cmap='RdBu')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 445
    },
    "id": "A1feLDbSZtXJ",
    "outputId": "a91d08b7-42cc-4671-b23c-ed01fac112f2"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report CNN + LSTM Model:\n",
      "               precision    recall  f1-score   support\n",
      "\n",
      "           0       0.92      0.96      0.94      2176\n",
      "           1       0.96      0.93      0.94      2324\n",
      "\n",
      "    accuracy                           0.94      4500\n",
      "   macro avg       0.94      0.94      0.94      4500\n",
      "weighted avg       0.94      0.94      0.94      4500\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAk4AAAD+CAYAAADMDIqpAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3deZyWVf3/8dd7BlAUkE1JFncqSREFcUmD1FzKHTN3rcyfpZWZJqRfTVyoNLW++i0pcctyIRdSSkshdwURXECU3GAwEQFZE5n5/P64zuA9w3Df98DczPZ+Ph7Xg/s617nOda6Zm/v+zDnnOkcRgZmZmZkVVtbYFTAzMzNrLhw4mZmZmRXJgZOZmZlZkRw4mZmZmRXJgZOZmZlZkRw4mZmZmRXJgZOZNThJIWmHEpV9oqRHcva/KOkNSUslHSnpb5JOLcF1fyfpfxq6XDNrXhw4Wasl6QRJk9MX7nvpC3efdOxn6cv/2Jz8bVLaNmn/lrQ/OCfPDpIafHK0tQUiktpJ+pWkOek+3pZ0XTq2NGerkrQiZ//EnHv8Ya0yf5jSf5anPltKuin93JZIek3SpZI2beh7ry0i7oiIA3OSRgLXR0SHiLg/Ig6JiFvX5xqSTpP0ZK3rnhkRl61PuWu51s8kfZJ+L4skPS1pr4a+Tj3q8sfGuLZZc+HAyVolSecC1wFXAj2ArYD/A47IybYAuFRSeZ6iFgCXr2MdfpYvOCnSCGAQMBjoCAwFpgCkQKJDRHQA3gUOy0m7I53/OnBKrTJPTelrq3dX4BmgPbBXRHQEvgJ0BrZfz/tZF1sDrzbCdRvSXen31B2YANyzoSsgqc2GvqZZc+TAyVodSZuRtVKcFRH3RsSyiPgkIv4aEefnZP07sBI4KU9xtwL9JQ0pYZXz2R24LyLmRubtiLitHudPAjaR9AWA9O/GKX1tzgWWACdFxNsAETE7In4YES/Vzizpa5JelLRY0uzcYFHSxpL+KOnD1NoySVKPdOw0SW+mFq23JJ2Yk/5kev1vYDvgr6nFZiNJEyWdnnON70iakcqZLmm3lD5c0r9z0o9K6TsCvwP2qm4FSum3SLq8VrmzJC2QNE5Sz5xjIenM1IW4SNINklTolxERq4A7gF6SNk9lbZbTulch6fLqYD79LJ6SdL2kj1LL3/459eiZ6rYg1fU7Ocd+Jmls+vkvBs4Efgp8I933tEL1NWuNHDhZa7QXWXBwX4F8AfwPcImktmvJs5ys1eqKhqtevTwLnCvpe5J2LubLuQ6382mr06lpP58DgHsjoqrI8pel8jsDXwO+K+nInOttBvQBupF9ea9IXX6/AQ5JLVp7A1NrFxwR21OzNe3j3OOSvg78LF2/E3A48GE6/G9g33T9S4E/StoyImakejyTyuxc+7qS9gNGAccCWwLvAHfWynYoWWDbP+U7qNAPSlK7VNcPgYUp+RZgFbADsCtwIHB6zml7pHvpDlwC3JtaBUl1mgP0BI4Brkx1r3YEMJbsd3MT2Xv5rnTfuxSqr1lr5MDJWqNuwPz0131eETEO+ICaX1S13QhsJemQBqpffYwCfgGcCEwGKlT/gdF/BI5PweFxaT+fbsB7xRYeERMj4uWIqEotUn8GqlvoPknl7RARlRHxQkQsTseqgJ0ktY+I9yJiXbrjTgd+GRGTUovcrIh4J9XrntRSVxURdwFvkHV5FuNEYExETEnB2giyFqptcvL8PCIWRcS7ZN1vA/KUd2xq2VoBfAc4JiJWpda3rwLnpJbRecC1ZL+navOA61Kr6V3ATOBrkvoAXwQuiIj/RsRU4A/U7Jp9Jo0Lq4qIFUXeu1mr5sDJWqMPge71GNNxEXAhWSvVGtIX52Vpy0vSg6nrZhEwHBhevS/pwSLrk3vtyoi4ISK+SNZqcAUwJnU3FVvGu8AsstaGNyJidoFTPiRrZSmKpD0kTZD0gaSPyFpzuqfDtwMPA3dKmivpl5LaRsQy4Bsp73uSHpL0+WKvmaMPWWtMXfU6RdLUnN/HTjn1KqQnWSsTABGxlOzn0isnz39yXi8HOuQp7+7UstUDeAUYmNK3BtqS/Qyq63kjsEXOuRVRc7X2d1L9egILImJJrWO5dSz0uzazWhw4WWv0DPAxcGShjAAR8Q+ywOJ7ebLdTBa4HF2grEMjonP6kvw5WatE57QdWlTt1172ioi4gayLp189T78N+HH6t5B/AkdJKvbz40/AOKBPRGxGNn5Iqc6fRMSlEdGPrDvuUFKLSEQ8HBFfIQvSXgN+X4/7qTabOgasS9o6lXc20C39Pl6prhdZN20+c8mCmuryNiVrOatYhzquFhHzgTOAn0naMtX/Y6B7zvukU0R8Iee0XrW6aLdK9ZsLdJXUsdax3DrWvs8GfyLUrKVx4GStTkR8BFwM3KBs3p9NJLWVdIikX67ltAuBn+QpcxXZ+JILGr7Gq7VLg6mrt3JJ50gaKqm9sukSTiV7uu7FepZ9F9nYmbuLyHsN2XihW1MAgqRekq6R1L+O/B3JWj7+q2zqhhOqD0j6chqbVQ4sJuu6q5LUQ9IRKSD5GFhK1nVXX38AzpM0UJkdUp03JQsSPkj1+CZZi1O194HeacxRXf4MfFPSAEkbkbXWPVc9WH59RMRMsla4n0TEe8AjwK8kdZJUJml71XwYYQvgB+k9/HVgR2B8ajl8GhiV3i/9gW+Tvyv2fWCbegTFZq2O/3NYqxQRvyJ7Ouwisi/P2WStD/evJf9TwPMFiv0z9Rj7sw5eJRsDU719k6wL6Fdk3ULzgbOAYRHxZn0KTq1V/yxmnEtELCBrHfoEeE7SEuBR4COylrnavgeMTPkupmZw9hmywcmLgRnAv8i678rIfj9zyaZ8GAJ8tz73lOp6D1n35Z/IngS8H+gaEdPJfm7PkAULOwNP5Zz6GNnP+z+S5tdR7j/JHhz4C9nvfHtqjjtaX1cBZ0jagqwFrh0wnaw1cSw1u0qfA/qS/f6vIBsfVT0A/nhgG7Kf433AJanua1M9DcKHkqY0zK2YtSyq2TVuZmbNhaTTgNMjYp/GrotZa+EWJzMzM7MiOXAyMzOzJk/SwZJmpslch9dxfGtJj0p6SdlEuL1zjm0l6RFlk+FO16dLZ20r6blU5l15xjV+eh131ZmZmVlTlh4geZ1seac5ZKsbHJ/GK1bnuQd4MCJuTRO9fjMiTk7HJgJXRMQ/JHUAqiJiuaS7ySb0vVPS74BpEfHbfHVxi5OZmZk1dYOBWRHxZkSsJJsV/4haefqRPdgB2aSzRwBI6ge0SVPLEBFLU9AkYD+yBy4gW0Kr4DQ1DpzMzMysqetFzQlb51BzMleAaXw6l95RQEdJ3YDPAosk3ats3cyrUgtWN2BRzioSdZW5Bq+GnUebdu2jbfuOhTOa5bHFxsEmbfw3iq2fj1XOJp27QkSds1QKoB5LFa5tmIbqudxhg5TjeyptXWqVM3369PkRsXm9ClgPZZ16B6v+mzdPrPjwVSA30+iIGF3PS50HXJ+eNn2cbLLXSrJYZ1+ytR7fJZu37jTggXqWDzhwyquqzaZUbrdekzmbccPB7dlt03WZu9HsU1M37cV+Z/64sathLUD7TTZ5p3CuBlT5MW13PCpvlpVT/vDfiBiUJ0sF2RJK1XpTa6b+iJhLanFK45iGRcQiSXOAqdXz20m6H9gTGAN0ltQmtTqtUWZd/GewWanV8y9Ds7r4XWTNmcrK825FmAT0TU/BtSObcHZcjWtI3XNmvR9BFhhVn9tZUnUr237A9LTG4wTgmJR+KkW0QjlwMisxf+FZg3AAbs2W1jtwSi1CZ5MtRzSDbGHsVyWNlHR4yjYUmCnpdbIFs69I51aSdeM9Kullso/l6rUvLwDOlTSLbMzTTYXq4q46MzMzKxlJlLUtOD1SQRExHhhfK+3inNdj+fQJudrn/gNYYy3N1H03uD71cOBkZmZmpSNRVlx3XLPgwMnMzMxKqshxTM2CAyczMzMrGUmUtWnb2NVoMA6czMzMrITkFiczMzOzosiBk5mZmVlxGuipuqbCgZNZidW9YIKZWesgPDjczOqhai1rTZnVR1WVl+2xZspddWZmZmZFkihv4646MzMzs4Lkp+rMrF7cVWcNwF2+1pw5cDIzMzMrhsooc1edmRVLEn62ztZXeVlZY1fBbN0IVO4WJzMrkhq7AmZmjchjnMzMzMyKJbWorjq3/ZqVWNZVZ7Z+/D6y5qysTHm3Ykg6WNJMSbMkDa/j+NaSHpX0kqSJknrnHKuUNDVt43LSb5H0Vs6xAYXq4RYnMzMzKxlJqMjgKE8Z5cANwFeAOcAkSeMiYnpOtquB2yLiVkn7AaOAk9OxFRGxtqDo/IgYW2xd3OJkZmZmJSUp71aEwcCsiHgzIlYCdwJH1MrTD3gsvZ5Qx/EG4cDJzMzMSkdQ3qYs71aEXsDsnP05KS3XNODo9PoooKOkbml/Y0mTJT0r6cha512RuveulbRRoYo4cDIzM7OSyRb5Vd4N6J4Cm+rtjHW41HnAEEkvAkOACqAyHds6IgYBJwDXSdo+pY8APg/sDnQFLih0EY9xMjMzsxISZYW74+anwGZtKoA+Ofu9U9pqETGX1OIkqQMwLCIWpWMV6d83JU0EdgX+HRHvpdM/lnQzWfCVl1uczMzMrHQEZW3K8m5FmAT0lbStpHbAccC43AySukuqLmwEMCald6nugpPUHfgiMD3tb5n+FXAk8EqhirjFyczMzEpGougpB9YmIlZJOht4GCgHxkTEq5JGApMjYhwwFBglKYDHgbPS6TsCN0qqImsw+nnO03h3SNqcrEdxKnBmobo4cDIzM7OSUgP0b0XEeGB8rbSLc16PBdaYViAingZ2XkuZ+9W3Hg6czErMq9pbQwi/j6yZkkR5ecsZGeTAyazE/HVnDcGBkzVn6zsBZlPiwMnMzMxKyoGTmZltUG5vsuZKwl11ZlYP7mKxBuCuOmu+1n+tuqbEgZOZWTNQ5HpeZk1PA0xH0JQ4cDIrsWzGXLcW2PopYuZlsyZJQFl5y3n/OnAyMzOz0nGLk5mZmVnxPMbJzIrmsSnWEPw+subKE2CamZmZ1YNbnMzMzMyKIEG5AyczMzOzwoRo18ZddWZmZmaFucXJWosD996Ja84/gbIycfP9T3DVzeNrHN9qy26MvuSbbN6lIwsWL+O0C39PxbyFAPT5TFd+d/Fp9OnRlQAOP/ta3nnvw0a4C2tsG23Xj80OOhYklk99iqVPP1LjePlmXel86MmUbdKBqv8uZ+H9N1O1ZBHlm3WlyzH/LxsUXV7OskkTWT7liUa6C2sKHnnkEc47/3wqKys57bTTOP+882ocf+fddznzzDOZP38+Xbp0YcxNN9G7d+/VxxcvXsyuu+3GYYcdxnXXXruhq99qiZYVOLWctjNrUGVl4tfDT+Kws69ll2EX8Y2D92DH7XrWyPOLHx3LHQ89zcBvXMIVo8dx+feHrT425rLTuebWv9N/2EXsfdJlzFu4ZEPfgjUFEpsdchwf/vl65v1uJO2/sDttun+mRpZO+x/N8pef44PfX8GSJ8bTab8jAahc8hHzb7mKD/5wJfPH/JIOex9EWYfNGuMurAmorKzknB/9iAfuv58Xp0zhnnvuYcaMGTXyjBgxghNPOIFJzz/PT0eM4OJLLqlx/NKRI9lnn302ZLUNKBNs1KYs71YMSQdLmilplqThdRzfWtKjkl6SNFFS75xjlZKmpm1cTvq2kp5LZd4lqV3B+ynyvutN0hhJ8yS9UiDfOZJOSa+vkvRauun7JHVO6YNzbniapKNSeh9JEyRNl/SqpB/WUf6PJYWk7mn/UEkjG/6OW5bdd9qOf8+ex1sVH/DJqkrufvg5Dhs6oEaeHbfryYTnsw+uiZNe47Chu65Ob1NezqPPTQdg2YqPWfHflRv2BqxJaNtzG1Yt+IDKRfOhqpIVr05m48/uUiNPm8235OO3ZwKw8u2ZbPzZ/tmBqkqoXJUytclGmFqrNWnyZLbffnu23XZb2rVrx9ePOYYHH3ywRp7XXnuNIUOHAjBkyJAax6dMmcK8efM4YP/9N2S1DUCivCz/VrgIlQM3AIcA/YDjJfWrle1q4LaI6A+MBEblHFsREQPSdnhO+i+AayNiB2Ah8O1CdSlli9MtwMH5MkhqA3wL+FNK+gewU7rp14ERKf0VYFBEDEhl3pjOXQX8OCL6AXsCZ+X+ICX1AQ4E3s257EPAYZI2Wb/ba9l6bdGZOe8vWL1f8f5Cem7epUael16fzZH7DQTgyP12o1OH9nTdbFP6btWDRUuWc/fVZ/H8ny9h1Dlfb1Gzxlrxyjt2pnLxwtX7lUsWUt6xc408n7xfQfvPZUH5xp8bQNlG7VH7TQEo69SFzb9zIT1+cCVLn36EqqUfbbjKW5Myd+5cevfqtXq/V69eVMydWyPPzjvvzAMPPADAAw88wJIlS/jwww+pqqpi+IgRjLryyg1aZ8tkXXVlebciDAZmRcSbEbESuBM4olaefsBj6fWEOo7XrFc2Odp+wNiUdCtwZKGKlCxwiojHgQUFsu0HTImIVemcR6pfA88CvVP68pz0jUkLf0XEexExJb1eAswAeuWUfy3wE3IWCotsifGJwKHrfHMGwAXX3s2XBn6O5/98CfsO/Bxz3l9AZWUVbdqUsc+ufbng2rvZ66TL2K735pxyeOttHq/0qvZ5Lf7nX2i3dV82P/2ntNu6bxZoVVUBULV4IR/8/grm3XAxm/Tfk7JNOzZybRtPVfqZ2NqNuvJKnnjiCfbcc0+eePJJevbsSXl5OTfeeCMHHXRQjfFOtuFI0K5NWd6tCL2A2Tn7c6j5fQ8wDTg6vT4K6CipW9rfWNJkSc9Kqg6OugGLcuKLuspcQ2MPDv8i8MJajn0LuKt6R9IewBhga+DknButPr4NsCvwXNo/AqiIiGl1zLg7GdgXuLv2AUlnAGcA0HbTet5Oy1ExbxG9e3Rdvd+rRxfmfrCwRp73PljEsefdAMCm7TfiqP0H8tHSFcx5fyHTXp/NWxUfADBuwosM3nl7bsEDe1ubyiWLKO/0aUtleccuVC5ZVCNP1dKPWDh2NABquxHtP78r8fGKNfKs+mAu7frswH9fe7H0Fbcmp2fPnsypqFi9X1FRQa+ePdfIc9eddwKwdOlS7r//fjp37sxzzz/PU089xejRo1m2bBkrV66kQ4cOXH7ZZRv0HlorUVR3XHdJk3P2R0fE6Hpe6jzgekmnAY8DFUBlOrZ1RFRI2g54TNLLwDo1YTf24PAtgQ9qJ0q6kKwb7o7qtIh4LiK+AOwOjJC0cU7+DsBfgHMiYnHqhvspcPFarjsP6FnXgYgYHRGDImKQ2mxcV5ZWYfKrb7HDVj3Ypmd32rYp59iD9uDBiVNr5OnWucPqZSAu+NbXuPWBJ1ef27njJnTvkrUODN19R2a8WbNJ3VqHT+a+Q5uuW1DeuRuUldP+C4P47+sv1chT1n5TssZ86PDFg1g+7eksvWNnaNMWAG28Ce36bM+qD9/foPW3pmPQwIHMmjWLt99+m5UrV3LP2LF87Wtfq5Fn/vz5q1vmrrrqKk495RQAbrn5Zt54/XVmvvYao668khNOOMFB0wZWLuXdgPnV371pqx00VQB9cvZ7p7TVImJuRBwdEbsCF6a0RenfivTvm2S9TrsCHwKd09CfOsusS2O3OK0g63pbLUWKhwL7p261GiJihqSlwE7AZEltyYKmOyLi3pRte2BboLq1qTcwRdLgiPhPuuaK2mXbpyorqzjnF3/kof87l7KyMm594EmmvzmXS757JC9Mf5sH/zWVIYM+x2XfPwYieGLK6/xg1B8BqKoKLrjmLh7+3XlIYsqMt7np3n818h1Zo4gqPvr7nXQ7/vtQVsbyqU+zav57dBxyKCvnvsvHb7xEu60/mz1JF8HH787io79nLQZtu3+GTgd8+qTm0mf/yaoPWm8AXsfHYavSpk0brr3mGg47/HAqKys59ZRT6NevHyNHjmS33Xbj0EMP5fEnnuDiiy9GEvt88Ytcd911jV1t49OuuvU0CegraVuy4OY44ISa11F3YEFEVJGNkR6T0rsAyyPi45Tni8AvIyIkTQCOIRszdSrwQMH7KeV/xtR99mBE7LSW42cCvSPiorR/MHANMCQiPsjJty0wOyJWSdoaeAboTxYt3kr2gzonTz3eJhtcPj/t/xhoGxE/z1f/sk26R5vPHZ4vi1lBYw/amEEdWveXnq2/F9pvyVe+d35jV8NagPabbPJCRAzaUNf7TN8vxCnXrDEypoarDt+pYJ0kfRW4DigHxkTEFekp+ckRMU7SMWRP0gVZV91ZKVjaG7gRqCLrabsuIm5KZW5HFjR1BV4EToqIj/PVo2QtTpL+DAwl67ecA1xSXdEcfwNuz9m/HtgI+EdqKXo2Is4E9gGGS/qE7Ma/FxHzJe0DnAy8LKm6H+mnEVFzpsY1fZlPn9gzKzGR83yC2Top83QM1kwVOcapoPTdPr5W2sU5r8fy6RNyuXmeBnZeS5lvkj2xV7SSBU4RcXwRed6R9KGkvhHxRppHoa58t1MzwKpOf5LqwRH5r7NN9WtJPYD2EfFyofPMGoJnYrCGUMdDLmbNQgN11TUZjT3GCWA42SDxNzbQ9bYCfryBrmVmZtaqtbQlVxo9cIqImcDMDXi9SRvqWmaAZ7y2BuF3kTVbXuTXzOqj5XxcWKNyAG7NlJC76szMzMyKIbc4mZmZmRXHY5zMzMzMiiSJduXuqjMzMzMrSkuah8yBk5mZmZWMgPKWEzc5cDIzM7PSkaCtu+rMzMzMiuOuOjMrmlepM7PWzF11ZlYvVeHQydZfVVVVY1fBbJ1Ioo276szMzMyK4xYnMzMzsyKIljXGqeW0nZk1Ve6qswbgLl9rrqqfqsu3FVeODpY0U9IsScPrOL61pEclvSRpoqTetY53kjRH0vU5aRNTmVPTtkWhejhwMjMzs5IqV/6tEEnlwA3AIUA/4HhJ/Wpluxq4LSL6AyOBUbWOXwY8XkfxJ0bEgLTNK1QXB05mJaYW1ERtjae8zB/X1jwJUab8WxEGA7Mi4s2IWAncCRxRK08/4LH0ekLucUkDgR7AI+t7P/6faFZiDpvMrDXLuuqUdytCL2B2zv6clJZrGnB0en0U0FFSN0llwK+A89ZS9s2pm+5/VMRfug6czMzMrKTKpbwb0F3S5JztjHW4zHnAEEkvAkOACqAS+B4wPiLm1HHOiRGxM7Bv2k4udBE/VWdWYu6qs4bg95E1V0U+VTc/IgblOV4B9MnZ753SVouIuaQWJ0kdgGERsUjSXsC+kr4HdADaSVoaEcMjoiKdu0TSn8i6BG/LV1EHTmZmZlYy1V1162kS0FfStmQB03HACTWvo+7AgoioAkYAYwAi4sScPKcBgyJiuKQ2QOeImC+pLXAo8M9CFXFXnZmZmZVQ/m668iJaUyNiFXA28DAwA7g7Il6VNFLS4SnbUGCmpNfJBoJfUaDYjYCHJb0ETCULyH5fqC5ucTIzM7OSaagJMCNiPDC+VtrFOa/HAmMLlHELcEt6vQwYWN96OHAyMzOzkmmgrromw4GTmZmZlVRLerah3mOcJHWR1L8UlTEzM7OWpwzl3ZqTolqcJE0EDk/5XwDmSXoqIs4tYd3MzMysmRNQ5HJ0zUKxt7JZRCwmmx/htojYAzigdNUyMzOzFkFZV12+rTkpNnBqI2lL4FjgwRLWx6zF8ar21hDC7yNrplSgm65FdtWRrTL8MPBUREyStB3wRumqZdZy+OvOGoIDJ2vOWlJXXVGBU0TcA9yTs/8mMKxUlTIzM7OWo3m1KeVXVAwo6bOSHpX0StrvL+mi0lbNzMyqub3JmqvqCTDzbc1JsY1nvydb9+UTgIh4iWydGDMrxF0s1gDcVWfNWbnyb81JsWOcNomI52utzr2qBPUxM7M6qJn9VW5WLXtyruW8f4sNnOZL2p7UWizpGOC9ktXKrAXJmqHdWmDrp7l1Z5jlKmtBb99iA6ezgNHA5yVVAG8BJ5WsVmZmZtYiiObXHZdPsU/VvQkcIGlToCwilpS2WmZmZtZStKSuumKfqvuhpE7AcuBaSVMkHVjaqpm1DC3pA8Maj99H1mwp66rLtzUnxT5V96205MqBQDfgZODnJauVmZmZtQjVXXUt5am6YgOn6tv6Ktlada/SsuazMjMzsxKRlHcrsoyDJc2UNEvS8DqOb53mnHxJ0kRJvWsd7yRpjqTrc9IGSno5lfkbFVGZYgOnFyQ9QhY4PSypI1BV5LlmZmbWSmUTYK5fV52kcuAG4BCgH3C8pH61sl1N1rjTn2ypuFG1jl8GPF4r7bfAd4C+aTu4UF2KDZy+DQwHdo+I5UBb4JtFnmtmZmatlihX/q0Ig4FZEfFmRKwE7gSOqJWnH/BYej0h97ikgUAP4JGctC2BThHxbGQzzN4GHFmoIsVOR7AXMDUilkk6CdgN+HWR5zZbHdoEu3dZCRLK6ZlUgV7K7Of/6bw9kV5Ln8aphcrIyqnZqJeVU7NZs2BdiBozV1fXxfe04e7pkTcXMOsz3eu6EapyZ4OO6uuLGp8jBT5UIqLm5ORpp6z2n3EFyqmqWrMu2aDOnPMKfcD5ntZezHre07IFH9LxhRcgqt9hUFaW817NZhnMe/2qqqoa+5D9LGr8/yvL//d0VVVVjVnMq1+X555XRPdLZWVlbuV8TznX3mD3tKHkv+Vq3SVNztkfHRGjc/Z7AbNz9ucAe9QqYxpwNFl8chTQUVI3YCHwK7JplA6oVeacWmX2KlTRYgOn3wK7SNoF+DHwB7LIbEiR5zdLGy1ZSt8JTzZ2NayZm7j/Afx25tLGroY1c8ftuw3nDhzY2NUwqzdFoMJLBs2PiEHreanzgOslnUbWJVcBVALfA8ZHxJyGeDq12MBpVUSEpCOA6yPiJknfXu+rm5mZWYunqvVepa0C6JOz3zulrRYRc8lanJDUARgWEYsk7QXsK+l7QAegnaSlZC1TvfOVWZdiA6clkkaQNXN9SVlfRtsizzUzM7NWKyDW+3mySUBfSduSBTfHASfkZpDUHVgQ2fiJEcAYgIg4MSfPacCgiBie9hdL2hN4DjgF+N9CFSm2s/MbwMfAt/z7H9MAABvkSURBVCPiP2RR2VVFnmtmZmatWUT+reDpsQo4G3gYmAHcHRGvShop6fCUbSgwU9LrZAPBryiiZt8jG340C/g38LdCJxS75Mp/gGty9t8lG+NkZmZmtnbRIC1ORMR4YHyttItzXo8FxhYo4xbglpz9ycBO9alHsUuu7ClpkqSlklZKqpT0UX0uZNZqFfHXlJlZS6aqVXm35qTYMU7Xk/Un3gMMIusH/GypKmXWkgQOnGz91Z72wqz5aJgWp6ai6AkdImIWUB4RlRFxM0XMrmlmZmatXJAFTvm2ZqTYFqflktoBUyX9EniPegRdZmZm1loFqmxe3XH5FBv8nAyUk41oX0Y2l8KwUlXKzMxq8lA5a9ZaW4tTRLyTXq4ALi1ddczMzKxFKXLKgeYib+Ak6WVY+8jWtAKxmeWRzRfbvP6isqZHxSwhb9ZENbcn5/Ip1OJ0NNkkUrNrpfcB/lOSGpmZmVkL0rqeqrsW+Cgi3sndgI/SMTMzM7P8WtEYpx4R8XLtxIh4WdI2JamRmZmtQbirzponRbSqrrrOeY61b8iKmJmZWQtV1bxalfIp1FU3WdJ3aidKOh14oTRVMjMzs5ajwAK/zeyJu0ItTucA90k6kU8DpUFAO+CoUlbMzMzMWoBoRU/VRcT7wN6Svsynqwc/FBGPlbxmZmZm1gK0rKfqip0AcwIwocR1MTMzs5aoBQVOXm/OzMzMSieCWPVJ3q0Ykg6WNFPSLEnD6zi+taRHJb0kaaKk3jnpUyRNlfSqpDNzzpmYypyati0K1aPYRX7NzMzM1kFAVeV6lSCpHLgB+AowB5gkaVxETM/JdjVwW0TcKmk/YBTZWrvvAXtFxMeSOgCvpHPnpvNOjIjJxdbFLU5mZmZWOkEWOOXbChsMzIqINyNiJXAncEStPP2A6jHYE6qPR8TKiPg4pW/EesY+DpzMSq2ZPWprTVP4fWTNVEQQn3ySdytCL2ou/zYnpeWaRrZUHGRP/neU1A1AUh9JL6UyfpHT2gRwc+qm+x9JBWeadeBkVmKx9nWyzerB7yNrrqKYFqfukibnbGesw4XOA4ZIehEYAlQAlQARMTsi+gM7AKdK6pHOOTEidgb2TdvJhS7iMU5mZmZWOhFE4e64+RExKM/xCqBPzn7vlJZzmZhLanFKY5mGRcSi2nkkvUIWJI2NiIqUvkTSn8i6BG/LV1G3OJmZNQNub7Lmq0GeqpsE9JW0raR2wHHAuNwMkrpLqo5rRgBjUnpvSe3T6y7APsBMSW0kdU/pbYFDgVcKVcQtTmZmzYEjJ2uuqgeHr08REasknQ08DJQDYyLiVUkjgckRMQ4YCoySFMDjwFnp9B2BX6V0AVdHxMuSNgUeTkFTOfBP4PeF6uLAyazEslXt/a1n66ngkFWzpioaZJHfiBgPjK+VdnHO67HA2DrO+wfQv470ZcDA+tbDgZNZqcmBk62/Ih72MWuaIohPVjZ2LRqMAyczMzMrnVj/CTCbEgdOZmZmVlLRAF11TYUDJzOzZkAe5GTNVQSscledmZmZWUFBuMXJzMzMrCgNMB1BU+LAyczMzErHT9WZmZmZFath5nFqKhw4mZmZWWm5q87MzMysCBFUFbceXbPgwMnMzMxKJ4KodFedmZmZWUEROHAys+JFtJwPDGs8fh9ZsxVB1SerGrsWDcaBk5mZmZVUS2pxKmvsCljT1e+gIfzstUcZ+cZEDrrgu2sc77pVL8755x1cNO1vnDvhTjr3+gwAvXfpx0+evpeLX3mEi6b9jYHHHrqhq25NyIF778Qr913J9AdGcf43v7rG8a227Mbff3ceL9x1Kf/4/U/otUWX1cf6fKYrD/3fubz0l8uZ9pfL2XrLbhuy6tbEPPLII/TfZRe+sNNOXHX11Wscf+fddznkq19l98GDOfCgg5gzZ06N44sXL2b7HXbgnB/9aENV2YCIoKqyMu/WnJQscJJ0sKSZkmZJGp4n33WSvpRe35HOeUXSGEltU/rnJT0j6WNJ5+Wc+zlJU3O2xZLOyTn+fUmvSXpV0i9T2s6SbinVfbcUKivj+BtGcv0hp3Fpv6+w+/GHs+WOO9TIM+zqn/Lsbfdy+S6H8NDIX3PkqJ8AsHL5Cm455VxG7nQg/3vwqRx73cW036xTY9yGNbKyMvHr4Sdx2NnXssuwi/jGwXuw43Y9a+T5xY+O5Y6HnmbgNy7hitHjuPz7w1YfG3PZ6Vxz69/pP+wi9j7pMuYtXLKhb6HJiGjsGjSuyspKzvnRj3jg/vt5ccoU7rnnHmbMmFEjz4gRIzjxhBOY9Pzz/HTECC6+5JIaxy8dOZJ99tlnQ1bbYHVXXb6tGIXiCklbS3pU0kuSJkrqnZM+JcUJr0o6M+ecgZJeTmX+RlLBRSFLEjhJKgduAA4B+gHHS+pXR75uwJ4R8XhKugP4PLAz0B44PaUvAH4A1PgTIyJmRsSAiBgADASWA/elsr8MHAHsEhFfqD43Il4GekvaquHuuOXZZvAA5s16h/lvzabyk0+YdOdf6X/EgTXybNmvLzMfexqAmROeYZcjvgLAvDfeYt6stwH46L15LJn3IR0377pB629Nw+47bce/Z8/jrYoP+GRVJXc//ByHDR1QI8+O2/VkwvPZF+DESa9x2NBdV6e3KS/n0eemA7Bsxces+G/LmX3Y6mfS5Mlsv/32bLvttrRr146vH3MMDz74YI08r732GkOGDgVgyJAhNY5PmTKFefPmccD++2/IalsSlVV5t0KKjCuuBm6LiP7ASGBUSn8P2CvFCnsAwyVV/wX3W+A7QN+0HVyoLqVqcRoMzIqINyNiJXAnWRBT2zDg79U7ETE+EuB5oHdKnxcRk4B8E0HsD/w7It5J+98Ffh4RH1eXkZP3r8Bx63ZrrUOXXj1YOHvu6v1Fc96jS68eNfLMmTaDXY/O3mMDjjqI9p06smnXzjXybLP7LpS3a8sH/36H1qo1r2rfa4vOzHl/wer9ivcX0nPzLjXyvPT6bI7cbyAAR+63G506tKfrZpvSd6seLFqynLuvPovn/3wJo875OmVlrfdnWcQfwi3a3Llz6d2r1+r9Xr16UTF3bo08O++8Mw888AAADzzwAEuWLOHDDz+kqqqK4SNGMOrKKzdonS0TkS3ym28rQjFxRT/gsfR6QvXxiFhZHQsAG5FiH0lbAp0i4tkUd9wGHFmoIqUKnHoBs3P256S02r4IvFA7MXXRnUxOUFWE44A/5+x/FthX0nOS/iVp95xjk4F961G21eEv511B3yF78NMpD/HZIXuycM57VOX85dDpM5tz2u3XcNs3zydacz9DK//CK+SCa+/mSwM/x/N/voR9B36OOe8voLKyijZtythn175ccO3d7HXSZWzXe3NOObwVd7P4bVTQqCuv5IknnmDPPffkiSefpGfPnpSXl3PjjTdy0EEH0bt378auYusUULlyVd6tCMXEFdOAo9Pro4COqWcLSX0kvZTK+EVEzE3n5w6EW1usUkNjP1W3JfBBHen/BzweEU8UU4ikdsDhwIic5DZAV2BPYHfgbknbpahyHtBzjYKyss4AzgDoQHmRt9HyLKx4ny59Pv0Rde69JQsr3q+R56P35nHjsKyreKNNN2HXYQez4qPFAGzcsQNnP3Qz4y68mreee3HDVdyalIp5i+jd49Nu2l49ujD3g4U18rz3wSKOPe8GADZtvxFH7T+Qj5auYM77C5n2+mzeqsg+IsZNeJHBO2/PLRT1sWAtTM+ePZlTUbF6v6Kigl49e66R56477wRg6dKl3H///XTu3Jnnnn+ep556itGjR7Ns2TJWrlxJhw4duPyyyzboPbRaxU2A2V3S5Jz90RExup5XOg+4XtJpwONABVCZVSFmA/1TF939ksbWs+zVShU4VQB9cvZ7p7TaVgAb5yZIugTYHPh/9bjeIcCUiMj9Zp8D3Fvd7SepCuhOFqhtnK69hvSLGg2wuTZqtc0k70yaxhZ9t6HbNr1ZVPE+ux93GDed8IMaeTbt1oXlCxYRERw84ns8PeZuAMrbtuXM+27k2dvuZcpf/tYY1bcmYvKrb7HDVj3Ypmd3KuYt5NiD9uCUETfWyNOtcwcWfLSMiOCCb32NWx94cvW5nTtuQvcuHZm/cAlDd9+RF6a/3Qh3YU3BoIEDmTVrFm+//TY9e/bknrFjueXmm2vkmT9/Pl27dqWsrIyrrrqKU085BaBGvttvv50Xpkxx0LQBBRTTHTc/IgblOV4wrkitSEcDSOoADIuIRbXzSHqFrNfpqVTOWsusS6m66iYBfSVtm1qDjgPG1ZFvBrD6US1JpwMHAcdH/WZ7O56a3XQA9wNfTuV+FmgHzE/HPgu8Uo/yW52qykruOvtifvDwbfxsxj954e4HeW/6Gxx26Y/of9gBAHxu6J5cOvMxLp35GB17dOdvV2StBgOP/Rp9vzSYvU47hgtfHM+FL46n9y5rPBtgrUBlZRXn/OKP2ZQC917B2EcmMf3NuVzy3SM5dEg2SHzIoM/xyv1X8ur9V7JFt06M+kM2oLeqKrjgmrt4+HfnMeXukUhw073/aszbaVSteawcQJs2bbj2mms47PDDGbDrrgw7+mj69evHyJEjVw8Cf/yJJ+i/yy7s3L8/8+bN44ILLmjkWhuQPVW3clXerQgF4wpJ3SVVxzUjgDEpvbek9ul1F2AfYGZEvAcslrRnepruFOCBQhVRqcaeSPoqcB1QDoyJiCvqyLMv8P8i4qS0vwp4B6h+5vjeiBgp6TNk45I6AVXAUqBfRCyWtCnwLrBdRHyUU3Y7sh/aAGAlcF5EPJaOXQ88HBF/zXcPm2ujGMaW6/wzMAN4Zv8DmPFhy5n8zRrHcftuw+hfnN/Y1bAWoP0mm7xQoHWnQfX/TPcYd+rX8ubZ9pe3FaxTXXGFpJHA5IgYJ+kYsifpgqyr7qyI+FjSV4BfpXQB11d3A0oaBNxC9iT/34DvR4HAqGRjnCJiPDC+QJ4nJI2S1DkiFkVEnfWJiP9Qszkt99gyYI1Z8dKo+5Nqp0vaCBgEnFP7mJmZmTWsoGEW+a0rroiIi3NejwXWGLsUEf8A+q+lzMnATvWpR2MPDgf4MbAVsKhQxgayFTA8IlrOwjlmZmZNVQRVn+SbTah5afTAKSKe28DXewN4Y0Ne08zMrNWKlrVWXaMHTmZmZtaSNUxXXVPhwMnMzMxKJiKoLHI9uubAgZOZmZmVjrvqzKxeWvNyM2ZmAVHZcj4HHTiZlVjQcj4wrPHUb05gs6Yj66qrbOxqNBgHTmZmZlZSVW5xMjMzMyssPMbJzMw2NA+Vs2YrgsqVDpzMzMzMCnOLk5nVR7ZYd8v50LDGoTI1dhXM1kkAVVUtp8nUgZOZmZmVTgSVK/1UnZmZmVlB4XmczMxsQxPuqrNmqoUFTmWNXQEzMzNrwdIEmPm2Ykg6WNJMSbMkDa/j+NaSHpX0kqSJknqn9AGSnpH0ajr2jZxzbpH0lqSpaRtQqB5ucTIzM7OSCdZ/AkxJ5cANwFeAOcAkSeMiYnpOtquB2yLiVkn7AaOAk4HlwCkR8YaknsALkh6OiEXpvPMjYmyxdXGLk5mZmZVOBFFZlXcrwmBgVkS8GRErgTuBI2rl6Qc8ll5PqD4eEa9HxBvp9VxgHrD5ut6OAyczMzMrmQioXFmVdytCL2B2zv6clJZrGnB0en0U0FFSt9wMkgYD7YB/5yRfkbrwrpW0UaGKOHAyMzOzkoqqqrwb0F3S5JztjHW4zHnAEEkvAkOACmD1ACpJWwK3A9+MT1fNHgF8Htgd6ApcUOgiHuNkZmZmpRNRzBin+RExKM/xCqBPzn7vlJZzmZhLanGS1AEYVj2OSVIn4CHgwoh4Nuec99LLjyXdTBZ85eXAyczMzEomAqrWfwLMSUBfSduSBUzHASfkZpDUHViQWpNGAGNSejvgPrKB42NrnbNlRLwnScCRwCuFKuKuOjMzMyudyJ6qy7cVLCJiFXA28DAwA7g7Il6VNFLS4SnbUGCmpNeBHsAVKf1Y4EvAaXVMO3CHpJeBl4HuwOWF6uIWJ7NS87L21gDC7yNrtqJBJsCMiPHA+FppF+e8HgusMa1ARPwR+ONaytyvvvVw4GRWYoG/8Kwh+H1kzVMEfFLVchY6d+BkZmZmJRNAC1pxxYGTmVlz0IK+d6wVqmxBXc0OnMzMmoOW871jrUxVwMqqlvMGduBkVmLZqvYt50PDGokauwJm685ddWZWPDlwsvWXTTNj1vwE4a46MzMzs2J4cLiZmZlZkcJjnMzMbEOTBzlZM+YWJzMzM7MiZF11LSdycuBkZmZmJRO4q87MzMysKBHuqjMzMzMrmrvqzMzMzIoQBJ84cDIzMzMrrKXN41TW2BUwMzOzlqv6qbp8WzEkHSxppqRZkobXcXxrSY9KeknSREm9U/oASc9IejUd+0bOOdtKei6VeZekdoXq4cDJzMzMSqZ6Asx8WyGSyoEbgEOAfsDxkvrVynY1cFtE9AdGAqNS+nLglIj4AnAwcJ2kzunYL4BrI2IHYCHw7UJ1ceBkVmIRVY1dBWsB/D6y5qwy8m9FGAzMiog3I2IlcCdwRK08/YDH0usJ1ccj4vWIeCO9ngvMAzZXtgDkfsDYdM6twJGFKuIxTnmUbdyOTbbbJpuxtx6T9matjmu+E+q7SGfU2Xwp6lVMZAPz6ijF97SB7qnLxmLH3h3qOrPOc7Pr1+8HUnc59fu7qK4v5oari+9pfe+pnFXMmjWrxvVVVo+fRwRVVWv+PMrKyqjPf8CoqqrzPsrKy4uvC1BVWblGmu+JJndPDaGBJsDsBczO2Z8D7FErzzTgaODXwFFAR0ndIuLD6gySBgPtgH8D3YBFEbEqp8xehSqiuj/0DUDSB8A7jV2PJq47ML+xK2Etgt9L1hD8Pips64jYfENdTNLfyX4v+WwM/Ddnf3REjM4p4xjg4Ig4Pe2fDOwREWfn5OkJXA9sCzwODAN2iohF6fiWwETg1Ih4VlJ34NnUTYekPsDfImKnfBV1i1MeG/KN1VxJmhwRgxq7Htb8+b1kDcHvo6YnIg5ugGIqgD45+71TWu515pK1OCGpAzAsJ2jqBDwEXBgRz6ZTPgQ6S2qTWp3WKLMuHuNkZmZmTd0koG96Cq4dcBwwLjeDpO76tO97BDAmpbcD7iMbOF49nonIutwmAMekpFOBBwpVxIGTmZmZNWmpRehs4GFgBnB3RLwqaaSkw1O2ocBMSa8DPYArUvqxwJeA0yRNTduAdOwC4FxJs8jGPN1UqC4e42TrRdIZuf3QZuvK7yVrCH4fWak5cDIzMzMrkrvqzMzMzIrkwKkVkVSZ+nZfkXSPpE0aoMyRkg7Ic/xMSaes73Ws+aj1Pvtrzgy9DVX+2+kxYiQtbciybcPLeb9Ub9tI6iZpgqSlkq5v7Dqa5XJXXSsiaWlEdEiv7wBeiIhrco5XP5Jpts5qvc9uBV6PiCsKnFaf8t8GBkXE/NxrWfNU1+9Q0qbArsBOZPPwnF3nyQ1fF38GWkFucWq9ngB2kDRU0hOSxgHTJZVLukrSpLQY4v+rPkHSBZJeljRN0s9T2i1pYjIk/VzS9HTe1SntZ5LOS68HSHo2Hb9PUpeUPlHSLyQ9L+l1Sftu6B+GlcwzpJl4JW0v6e+SXkjvuc+n9B7p/TAtbXun9PtT3lclndGI92AbWEQsi4gnqTkh4hokfSF9bkxNnyt9U/opaX+apNtT2jaSHkvpj0raKqXfIul3kp4Dfrm296lZNU+A2QpJakO2UOLfU9JuZH/VvZW+oD6KiN0lbQQ8JekR4PNk6/7sERHLJXWtVWY3sinuPx8RsZbumduA70fEvySNBC4BzknH2kTEYElfTelr7f6z5kHZopz78+njvaOBMyPiDUl7AP9Htk7Ub4B/RcRR6Zzq1odvRcQCSe2BSZL+krt0grUY7SVNTa/fioij6nHumcCvI+IOZXP1lEv6AnARsHdqlaz+rPpf4NaIuFXSt8jed9XrkvVO+SslPUrd71MzwIFTa5P7AfUE2Rfa3sDzEfFWSj8Q6F/digRsBvQlC2RujojlABGxoFbZH5H9dXiTpAeBB3MPStoM6BwR/0pJtwL35GS5N/37ArDNOt+hNQXV77NeZPOt/EPZLL57A/fo0/W2Nkr/7gecAhARlWTvJYAfSKr+Eu1D9j504NTyrIiIAYWz1ekZ4EJJvYF7U7CzH3BPRMyHGp9Ve5FmlQZuB36ZU849KWjK9z41Axw4tTZrfEClD4dluUlkrUIP18p3UL6CI2KVssUT9yebhfVs6vdX2sfp30r8vmzuVkTEAGUPHzwMnAXcQraYZlFfkJKGkgXre6UWzolka1lZK5YC6UvS7ukR8afUxfY1YHzu0IJ6qv4MLKMe71NrnTzGyWp7GPiupLYAkj6bBmr+A/hm+jKkjq66DsBmETEe+BGwS+7xiPgIWJgzfulk4F9Yi5VaJ38A/BhYDrwl6esAylS/Rx4FvpvSy1Pr5GbAwhQ0fR7Yc4PfgDU5EXFfRAxI22RJ2wFvRsRvyJbK6A88Bnw9DR/I/ax6mmyZDoATyVrda5e/mLW/T80A/2Vva/oDWVfZFGXNUR8AR0bE35VNUT9Z0kpgPPDTnPM6Ag9I2pis1ercOso+FfhdCr7eBL5ZutuwpiAiXpT0EnA82ZfVbyVdBLQF7gSmAT8ERkv6NlmL43fJxt+dKWkGMBN4tq7yreVS9vRkJ6CdpCOBAyNieq1sxwInS/oE+A9wZRoXdwXwL0mVwIvAacD3gZslnU/2uba2z5+1vU/NAE9HYGZmZlY0d9WZmZmZFcmBk5mZmVmRHDiZmZmZFcmBk5mZmVmRHDiZmZmZFcmBk5mtF0kh6Y85+20kfZBmkM933oC0xM7ajg+S9JuGrKuZ2fpy4GRm62sZsFNaUw7gK0BFEecNAOoMnJStUj85In7QQHU0M2sQDpzMrCGMJ1v2ArLJLv9cfUDSppLGpFXsX5R0RFqQdSTwjbSy/Tck/UzS7ZKeAm6XNLS61UpSB0k3S3o5rW4/LM0yfoukV1L6jzb0TZtZ6+PAycwawp3AcWnm+P7AcznHLgQei4jBwJeBq8hmZL4YuCstn3FXytsPOCAijq9V/v8AH0XEzhFRvazGAKBXROwUETsDN5fq5szMqjlwMrP1FhEvkS3VczxZ61OuA4HhkqYCE8kW691qLUWNi4gVdaQfANyQc72FZMv2bCfpfyUdDCxen3swMyuGAyczayjjgKvJ6aZLBAzLWZx1q4iYsZYylq0lfQ0peNqFLBg7k2ydRTOzknLgZGYNZQxwaUS8XCv9YeD7adFoJO2a0peQLQ5djH8AZ1XvSOoiqTtQFhF/AS4CdlufypuZFcOBk5k1iIiYExF1TR9wGdmYppckvZr2ASYA/aoHhxco/nKgSxoIPo1srFQvYGLqAvwjMKJBbsTMLA9FRGPXwczMzKxZcIuTmZmZWZEcOJmZmZkVyYGTmZmZWZEcOJmZmZkVyYGTmZmZWZEcOJmZmZkVyYGTmZmZWZEcOJmZmZkV6f8DPTIqJWjtUI4AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 708.661x255.118 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "lstm_cnn_prediction = (model_CNN_LSTM.predict(x_test) > 0.5).astype(\"int32\")\n",
    "print('Classification Report CNN + LSTM Model:\\n', classification_report(y_test, lstm_cnn_prediction))\n",
    "\n",
    "plot_classification_report(classification_report(y_test, lstm_cnn_prediction), title='CNN + LSTM Classification Report', cmap='RdBu')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "id": "yXjPDtRVZtXJ"
   },
   "outputs": [],
   "source": [
    "from sklearn import metrics\n",
    "from sklearn.metrics import roc_curve, auc\n",
    "\n",
    "fprB, tprB, thresholdsB = metrics.roc_curve(y_test, model_batch.predict(x_test))\n",
    "fprMB, tprMB, thresholdsMB = metrics.roc_curve(y_test, model_mini.predict(x_test))\n",
    "fpr3L, tpr3L, thresholds3L = metrics.roc_curve(y_test, model3L.predict(x_test))\n",
    "fprBi, tprBi, thresholdsBi = metrics.roc_curve(y_test, model_Bi_LSTM.predict(x_test))\n",
    "fprCNN, tprCNN, thresholdsCNN = metrics.roc_curve(y_test, model_CNN_LSTM.predict(x_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 350
    },
    "id": "IqnADfg_ZtXJ",
    "outputId": "f8f73257-6ab8-48be-e501-e0f6c82c627f"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfkAAAFNCAYAAAAD7RaHAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzdeXxU9b3/8ddnlmSyQkIIO7KL7CKLiiIC4gLiwk572962V+tSe39t/VWttRa7Xe29P9pqa73WWq0kIKhgESsqgoooIjuC7FsggZA9k2SW7++PM0wSMgmTZTJZPs/Hw0fOnPnOmU9CzHvO93zP9yvGGJRSSinV9tiiXYBSSimlIkNDXimllGqjNOSVUkqpNkpDXimllGqjNOSVUkqpNkpDXimllGqjNOSVUi2SiDwuIv+Idh1KtWYa8ko1ExE5IiJuESkWkdMi8qKIJF7Q5moReV9EikSkQETeFJEhF7RJFpHFInIscKyDgcdptbyviMgDIrJLREpE5ISIvCoiwyP5/YYj8DOoCHwf50RkrYgMbsBxjojI1EjUqFRrpiGvVPO61RiTCIwCLgcePv+EiFwFvAOsBLoDfYHtwMci0i/QJgZ4DxgK3AQkA1cBucC4Wt7z98APgAeAVGAQ8AYwvb7Fi4ijvq8Jw5OBn0lPIAd4MQLvoVS7pCGvVBQYY04D/8IK+/OeBF4yxvzeGFNkjDlnjHkU2AQ8HmjzDaA3cIcxZo8xxm+MyTHGPGGMeevC9xGRgcB9wAJjzPvGmHJjTKkx5hVjzG8DbT4Qke9Wec23ROSjKo+NiNwnIvuB/SLyZxH53QXvs1JEfhjY7i4iK0TkjIgcFpEHwvyZlAJLgGGhnheRmSKyW0TyAzVfFtj/cuBn8magR+D/iohLRP4hIrmB9ptFpEs4dSjVlmjIKxUFItITuBk4EHgcD1wNvBqi+TLghsD2VOBtY0xxmG81BThhjPmscRVzOzAeGAJkAPNERABEJAWYBmSKiA14E6sHokfg/f9TRG682BsELl18Ddga4rlBgff9T6Az8BZWqMcYY/4NOEagl8QY8yTwTaAD0AvoBHwPcDf821eqddKQV6p5vSEiRcBxrK7pnwf2p2L9/3gqxGtOAeevt3eqpU1t6tu+Nr8J9Cy4gQ8BA1wbeG428IkxJgsYC3Q2xiwyxlQYYw4B/wvMr+PYPxaRfKwPPInAt0K0mQesNsasNcZ4gN8BcVgfjELxYH3vA4wxPmPMFmNMYX2+YaXaAg15pZrX7caYJGASMJjK8M4D/EC3EK/pBpwNbOfW0qY29W1fm+PnN4y1qlUmsCCwayHwSmD7EqB7oIs8PxDejwB1dZX/zhjT0RjT1Rgz0xhzMESb7sDRKjX4AzX1qOWYL2NdDskUkSwReVJEnBf/NpVqWzTklYoCY8x6rAFmvws8LgE+AeaEaD4Xa7AdwLvAjSKSEOZbvQf0FJExdbQpAeKrPO4aquQLHmcAs0XkEqxu/BWB/ceBw4HQPv9fkjHmljDrrU0W1gcIwLpjAKsr/mSo+owxHmPML4wxQ7DO9mdgjWdQql3RkFcqehYDN4jIyMDjh4BvBm53SxKRFBH5Jdbo+V8E2ryMFaQrRGSwiNhEpJOIPCIiNYLUGLMf+BOQISKTRCQmMChtvog8FGi2DbhTROJFZADwnYsVbozZitW78DzwL2NMfuCpz4AiEfmJiMSJiF1EhonI2Ib8gKpYBkwXkSmBM/IfAeXAxsDz2UC/841F5HoRGS4idqAQq/ve38galGp1NOSVihJjzBngJeCxwOOPgBuBO7Guox/Fus3umkBYY4wpxxp8txdYixVgn2F1+39ay1s9ADwNPAPkAweBO7AGyAH8P6ACKyj/TmXX+8UsCdSypMr35MM6ax4FHKbyg0CHMI8ZkjFmH/B14I+BY96KNdCuItDkN8CjgUsEP8bqjViO9fP5EliP9QFJqXZFrMtrSimllGpr9ExeKaWUaqM05JVSSqk2SkNeKaWUaqM05JVSSqk2SkNeKaWUaqMisaJURKWlpZk+ffpEuwyllFKqWWzZsuWsMaZzQ17b6kK+T58+fP7559EuQymllGoWInL04q1C0+56pZRSqo3SkFdKKaXaKA15pZRSqo3SkFdKKaXaKA15pZRSqo3SkFdKKaXaKA15pZRSqo2KWMiLyAsikiMiu2p5XkTkDyJyQER2iMjoSNWilFJKtUeRPJN/EbipjudvBgYG/rsL+HMEa1FKKaXanYjNeGeM2SAifepochvwkjHGAJtEpKOIdDPGnIpUTUoppVS9GAO+wrCb+/2GsjLTZG9f7vE36vXRnNa2B3C8yuMTgX0a8kqp6PF7wO8O/ZyvBLx5tb/Wmw++IursJC0/DvjrbgPgyQFPHtgTrcfGA4WfgqtP3a8LR+keEHuVYxvIfw/iB9frMD6/jQpPDAYhvyg1eCi/qTvkfH47eUWdcNi9F+wP/TpjhNPneuCKcQePL1L78QtKK8g+14MeyWfwGqGWw5J1biAuZyk2m5evToyjS8qRGm3sUvniPcevolvKoZDHOpXXjxiHmwpvXO2F1ZORIvzODxt1jFYxd72I3IXVpU/v3r2jXI1SbZDxga/04u38JVbwABivFVjiDN3WvQ98bvCcBUdS09XaGCf/DAmXgS3whzhvrfXVngRIvc7YoqZgfb1fklvUlXKP9T2XlidxOGcop/P6Yozgrkjg+NlBdE6+rdpr9p28gk7Jp7CJL+QxT+X1q3/tzWxrPdvnl3S9aJu6vu+qAe9yFtfz3SsZvHhsO6iwbYNafv7himbInwR6VXncM7CvBmPMc8BzAGPGjGm6fhClWhJPHuS8Unl2VVXJbrAnEzx9cR+0gtmeULNt+UmoOA0xXaqEWHLt7+svB1Pe+Ppbi4KPau7zFdXY5a8I8VoBmxPK82wYqp9K+hFyHXZ6JZXzSW4Cof5Q5RX0oaQ0ne5x5ewptAKhpLgndntZyFITHX7OVdgp81tn/XbAAPkeOwAGwVvcl4qCy0D8eN3dsDmK8JV1A3yIvQzjC/E7EkJOwSU19mXn19wXit9Wis0fj9+Rj99R82cZiq2iE/6YXPzOc2G1xzgRXxze+MNhNXd4OhJrumBSOuDxWx9E/X4o8/pJ7xBTedhyMEl+/H5DsstB756xNY7lI5bzPS92p+Bwhe5GcMQI9hjB7hRE0gAocvvom+7C6ajSc2MgwWVnSI/q/zbGGLZv/YKVr6+i5FwuAKMuvwL+9rewvueQNTX4lY23CrhfRDKB8UCBXo9XYfFV+YRs/FbgeXKs7scLGT+U7gvvTLJ4GzjDWM2x9Cvrqz0+jLb7AH/NkM17B+KHBNrsufhxGqKkyna4Z6ihPmBUZYx1Nn++W7c8C2K6QmyvEG29mIpsyu2DIPYScHSs9nTx56vw5BzBi51yYvBi/ftV4KQCJ3X0xuLzx+D2hNc7cDDOQ4HdilwB0mK9bDybwPkrnV6/g+zCrpQH+nQNBEMVINbbiYSyvoix9sV4O+L0dcRjLwj5foINV0VX7P7qvx8Ofxi/L03A5z3/b2ivEfDumJP4xUNJ7FFKXcfwiwcAm9+J116C115ywdH8VDhqXp7omTKMK/vczYShHUnpVL0nxy5C/y5x2Gx1/QuqCxUVFfHsn//M3r17AejVqxcLFy7k0ksv5QcP3N/g40Ys5EUkA5gEpInICeDngBPAGPMs8BZwC3AAKAX+PVK1qGbgc8P5P5v+CijeWvm4dH/NQCzdB0VbIK4/5H9gfS38FGLSQSo/SZuS3VYX2PmQ9IX+w9pYXp+TwtJOIE3TUeTzO8gr7oLddv6aY/4FLcbVsh2QckPNfRXZkDC08rE33wpPW83u8nK3h1JvOthiAcPpvN4cPlhBl67V/5cv3P0RRXl5uMUFfvAEAraUOOz4AEMeHXDiJV+SSTDVr1U7/R3wSzlgEOPA6U/GaysNPHeRDwzMvMjzzcMGdGvm9yyOr94F6/QIp/HQL90FF3y8MRjKPYa4GBs9U2NJiLVjtwmxTutDh8NpI9YpJCU76HmJC4dDSOkUQ0yMkJhs/Xs7HEJMjA0Y3hzfnmqAhIQESktLSUxM5M4772TixInYbI2/AU7MRQZItDRjxowxup580zHGUFEe+B0wfvC78fkhL9dL7hkPZ894OXvGS0W5n5KiUuIdp6B0r3U91psPjg7gDS94v8oaTWl5EkXuTsTHhndmWVpeRzezajV8gQ8CTUewm1jcMVlhto7FRgoiQqLLHjzLNMZgtwketx9HjI2kVDuCEOOwYbOB0279kRUbXDosgdhY63F5uZ/0zjHEx4c+T3LGCD17u2qczdodEjyGat+8Xi/r1q3jiiuuIDXVGrR46tQpkpOTSUio3gMjIluMMWMa8j6tYuCdugh/OVTkWOEbosvaGHDnn+Bcdh6r13bjqyO96N3lCHnFnck606Web5Ye+K9x6hvezhipczRtU6goN3TpFkO4b+Q9l4Upv7B702KAIpOOS85RLl58Ysdu/JRIHE68uHEhGGxU3h5TQMM/0DidJ7HZqg/08diLMFI5ejmuojuFcfsoc2Y3+H1CsfvjcMdaw2nE2PHYC/DbrG5gv3god+bU2kMyps9VPPuNP1S/XlkvemaqWp9du3aRkZHBqVOnOHToEHfffTcA3bo1fZ+ShnyU+f0GjyfwB9BfAWVH8bgLObn3C/y+Csq9Cew/lMz+fX7KfJ1ISwlcH6s4U+NYe05cid/vwOkow2mvHEhVWt4BqH6teffhYdUexziqd8NWeOOw2bx0iM8lOS6P2Bg33ToeJC35JK6kVPAWQtrt4Ei2ruM600N2G1+ootzQvWcsvfuGf80uLt5WZ9vsggqKy7y1Pl/V2SIPu44Vk+Cq+WHI7zfsOFYcHJRTtHEZFacPVmuTL8mcoRM+bJx0duFQzCUMMEeqtyGZs5KKdRXKFaKKGIKXMqp1zVYfsBQfUxl88YnP4YyJ0LX7Kq4dOIGn/+0PEX8fpdqj7Oxsli5dyrZt2wBIT09n/PjxEX1PDfkIKMz34Hafvz5dTkneac7k+Nmw3sfZM4aePQ07d9R1BBdwdchnTueeP4u+tNZXe7wuPN5Q4VLp5skH6T8wFr9xMXRIGU6nH5LGWgPPmuiU2RhDWS0TOWw9XMSz754gwWUnIdYK3GNnyygq85KaYH1YOJlXzum8crqmVLlGb+DLkyWkJFi/unkl4YV7w4wDe4jr5Rc4IH0u2iY+1sascV3Ynf0Upwq/aHxpIWhAK9Uyud1u/vnPf/LOO+/g8/mIjY1l5syZTJ06Fafz4idHjaEh30hlZT4+/TCfjL9l4Q9zYqK8EHeMOB1VbqExNjy+GGziY1DfbPKLEvF4bQzud4qBQzqRmOgHjHV/sjPNCmabdfZpswmdu8SQmFT9TDUu3n7B2bDVzen3G3IKKygt90EeQOUZfV6JF3eFNUAov9TLjqPFxFe5nrjzWDG7TpQwoEscIrBpfwGFbh+pCQ4MTRfA50IcJ9Sxe5nar8/6UlZgXOHdehMp7+5v3Os1xJVqnc6cOcPbb7+NMYYJEyYwa9YsOnbsePEXNgEN+QYoKvSycd1Z3l51ipKS2n+E6R2OBbdzCnoTH1tMalIuVw/7gM6J+4m159Nr9J0kpA20GiVebt2O1IQXn/1+w5nCCr46Vspf3j3JZwcKKPP46Rjv4EB2LbN61dPhnOrHuTCUY0Legy1USAxX+HdytdmCC+um5BLi+KLTHo666n/fdiTP6ZuShrVSbd/Jkyfp0aMHYE3iNm/ePAYMGEC/fs07iZCOrq+HP//3UbZuDj0qPCUhG4Pw7amP0Td9N7HOMhj0HHSeHd791PVQ4fVz4lw5Hq+fNz4/w85jRaQnx+CwV55lf7Q3j/QOMWz8KvxbznrFFGDKK2c9M8BJ6cZov7WQYJ4k098cozs5wTZnSaGfOUZvrLPoCpyMNHsDt19BMsXYLhhV/atOnfkirummfgyXhqtSKtLy8vJYvnw5n3zyCT/4wQ8YOXJko4+po+sjwOv187tfHKak2Ev2qVBTX4Erppgx/d9l+pR9dEqLse5h7vgHa5ITe8NDzF3hY8uhQmyBM/ptR4t44rVD+Oq5TsGFZ+pJLjt9OscxreIdOmV9TDeTQyzl9AkEdH1OhasGdc2pI6uOEk+pX9EBGshKqdbE4/Hwzjvv8M9//pPy8nIcDgdnztQcIN3cNOSrKC/z8/wfj7F9S93TMv7xP661ztSv2AoJP6v3+3h9hsM5bj49UMCB7FL+ti6LuBgbXTrE8tXpkrDDvGdqLCfOWd3a1yYeZVzBWirvRRYqcNCbU4w0X5JIKRQDxYGA7nH+Q4gDaP71ADTElVJtgTGGrVu3snTp0mCojx49mrlz55Ke3vjbjRtLQx7ruvXiXx9m767Q9zzfeeUf6ZO+h35ddxLjKA+E+2X1fp/n3jvBo0sPhnzO7fHXuJbdId4RnNv4xLkyrrk0hQdvvYSOCQ4SXQ6yFi+gNOddq/GFE6phhfnyuDggrd61hkODWinV3n344Ye8+OKLAPTo0YMFCxYwZMiQ6BZVhYY88L2Fu6o9TkuP4Tt3J3BJzuDqSyEmjoLRm+o8lt9vOFfi4Wt/3MXWI1aPgMtpw+c3eHw1xz9MGpLCyXPlTB2eyph+yfROs2bJGtYzAakyAC9r8QJKN7xL2QY4XeX1Db2+rQGtlFINY4wJ/n0eN24ca9euZdKkSUyaNAm7PcQaGlHU7kP+0P7qy2v+4hdldCuYD7kH4fy/lT0Jhr8FyWNDHuPkuTJe/vAUL204xdkiT43nL7xX/NfzB/Ct67rjsIceRZ+1eAEHd7xba831CXYNc6WUaho+n4/169fzwQcf8Mgjj+ByuXC5XPziF79oknnmI6Fdh3xBvoff/qyy+/zZ323Cduj71Rv1fgT6PBby9cYYFr91jN+sPFLrezw+ux/XXpZCl8Asap2TnMFPgFmLF1BaR5ifd7FQ1yBXSqnI+vLLL1myZAknT1pTOG/atIlJkyYBtNiAh3Ye8g9+b29we9KwV7EderLyybQ7od9vwVVzUJoxhhfWZfFw5oFq+6eNSGV47yTuv7FXcBa32lQN+IZ0uWuwK6VU5J05c4Zly5axZcsWANLS0pg3bx6jR4+OcmXhabch/9qSyivbA7t9wfxrn6p88vKPIemKkK9b/mk29/51b439m345jn7pFw/qhoa7hrpSSjWv9957j6VLl+L1eomJiWHGjBnceOONEZ+Ktim1y5D3+w1vr6q8f/HBO6wVgIjpBuMP1TrjXIXXXyPgVz04kisHhjc94XefmMhmTwn0rNk7oCGulFItS6dOnfB6vVx55ZXMmTOHlJSGzfsRTe0y5B//ceUk4k8snGVtdLgGRoa+Pn4ou5Qrf7a52r5VD47iyoEdwnq/YLiHoOGulFItw+HDhzlw4AA33HADACNHjmTRokX07NkzypU1XLsL+aJCL6ezKudF79LxGAz6C3T9Zo22Pr/hx//4ilc+Ol1t/9CeCWEFfKhwH+tM4PmfbWhg9UoppZpaQUEBK1as4KOPPkJEGDx4ML169UJEWnXAQzsLeWMMj//4q+Djxd+53toIEfDGGLp9r3oYd+0Yw8ZFY0l01f1j03BXSqmWz+v1snbtWt58803Kysqw2+1MmzaNtLTITCAWDe0q5O/5+i781roppCRkEx9bDAP+GLLtuj151R6/88hoRvVJqvXY97/8AB/u/7jGfg13pZRqebZv305mZibZ2dkAjBo1innz5tGlS5coV9a02k3Il5f5gwEP8NM5/2ZtdPtuyPbzf78zuJ3z3HUXPf6FAa/hrpRSLdemTZvIzs6mW7duLFiwgGHDhkW7pIhoNyH/5a7i4PbTd11jzUE/ZmfIkfR/X58V3P7h9LoXb8lavICfntwOgVvh1qQOovt/ZjRR1UoppZpCaWkphYWFdO3aFYA5c+bQr18/rr/+ehyOthuFbfc7u8DeQMh365xnBTxA/MCQbRetOBTc/snMPiHbVLvuHgj4sc4EDXillGpB/H4/GzZs4LXXXiM1NZXHHnsMm81GampqcBR9W9ZuQv7gV9Yc9Ymus9aOHvfX2raozOrXf+4/Lqu2SMx5oQbW6a1wSinVsuzbt4+MjAyOHTsGQPfu3SkuLiY5OTnKlTWfdhPyZ7IrALi0d2Aq2vjQS8W6Kyov3I/tX/0X4cLBdXrdXSmlWp7c3FyWLVvG5s3W/CapqanMnTuXsWPHhjxxa8vaRcgXFXopLbHCe3Dn162dMV1Dtn3qzaPB7R6prmrPVQ340W43zy/a0sSVKqWUagyv18uvf/1r8vLyiImJ4eabb+amm24iNjY22qVFRbsI+T07Kgfd9esSGDXfYUKNdn6/4el/HQ95jO8+MTG4veLEMeJHTG3aIpVSSjWIMQZjDDabDYfDwS233ML+/fuZM2cOnTp1inZ5UdUuQv74ETcAQwZm47B7odMMcNScb/6Xrx8Obr/+o5HVnjt/DX602038iKk6wE4ppVqAo0ePkpGRwciRI7n55psBmDx5MlOmTIlyZS1Duwj5zZ8UAOApDQy6Kzsast3b284GtydcWvkh4P6XHwhu/+2pPRGoUCmlVH0UFhby+uuvs2HDBowx5OXlMW3aNOx2e7u77l6XdhHyKakO8nI9jOwdWICm8+yQ7Xp2cnEg281TX6t+a935a/Gj3e6I1qmUUqpuXq+X999/n5UrV+J2u7Hb7UyePJnbbrsNu90e7fJanHYR8of2W+Hcr+sOa0f6vJDtPghMZdsztXKARtWz+J/mnqnxGqWUUs0jPz+fp556ilOnTgEwbNgw5s+fT/fu3aNcWcvVLkL+vPjYImsj9pIaz+0+UXVwXnxwu+pZvA62U0qp6OnQoQPx8fGkp6czf/58Ro4cqV3zF9EuQt5uF3w+Q3qH4+DoFHIq24eWVK4x3zfdmsGu6oj6X/UYqYPtlFKqGbndbv75z39y3XXXkZ6ejohwzz33kJSUhNPpjHZ5rUK7CPlqbKHvlSwOzHJ39aDKdeLPj6jX6WqVUqr5+P1+Pv74Y1asWEFhYSGnT5/m+9//PmBNbKPC1/5C3hV6wRmvzwDwvak9gepn8TqrnVJKNY8DBw7wyiuvcPSodRdU//79mTFjRpSrar3aX8h3+UaNXXklHvadOj+3vZ2sxQuqncUrpZSKrLy8PF599VU2bdoEQMeOHZkzZw5XXnmlXndvhPYX8p3vrLHr0v+zMbg9tn8Hju94F3paZ/x6Fq+UUpFXVlbG5s2bcTgc3HTTTdxyyy24XK6Lv1DVqf2FPLXfR/mNid3IfeZrzViLUkq1T8YY9u7dy+DBgxERunXrxr//+78zcOBAOnfuHO3y2ox2GPK1uz/nF5TuepdfddJfMKWUipQTJ06QkZHBl19+yb333suYMWMAuPrqq6NcWdvTTkLe1PrMX98/Gdwu2/UeDuCLOOsWumsH1lzERimlVMMUFxfz+uuv88EHH2CMISEhAa/XG+2y2rR2EvJV2BOrPXw480Bw24Gv2nNP/9sfmqUkpZRqy3w+Hx988AFvvPEGJSUl2Gy24FS0iYmJFz+AarD2FfJxA6tNhGNM5Rn++z+7An6OdtUrpVQT++CDD3jllVcAuOyyy1iwYAE9e/aMclXtQzsJeQPUvAXj58sPBbeH9UrkANpVr5RSTcHj8QRnpbv22mv54osvmDJlCpdffrneEteMbNEuINI8FX58vsAvlHt/teeeXXui1tdpV71SStVfWVkZK1as4KGHHqK01Jp/JCYmhgcffJDRo0drwDeziIa8iNwkIvtE5ICIPBTi+d4isk5EtorIDhG5palryD5VHty2d61cYja/xBPcznxgOFmLF2hXvVJKNZDf72fjxo08/PDDrF69mry8PLZv3x7tstq9iHXXi4gdeAa4ATgBbBaRVcaYPVWaPQosM8b8WUSGAG8BfZqyjuJCa+Rm99SDSEXlmfvKzyuXjb1+aAoH/+ddvghMgKNd9UopFb5Dhw6RkZHBwYMHAejbty8LFy6kf//+Ua5MRfKa/DjggDHmEICIZAK3AVVD3gDJge0OQFZTF5G9ewVwBWcKesCQzOD+Tw8UAJCW5KzRfaRd9UopFZ5Vq1bxxhtvANZSsLNnz+aqq67CZmvzV4NbhUiGfA/geJXHJ4DxF7R5HHhHRL4PJABNvmC7HWsN+eF9PgLn2OD+co8fgGsHdyRr8YKmflullGoXBgwYgMPh4IYbbmDGjBnEBQYvq5Yh2qPrFwAvGmP+W0SuAl4WkWHGGH/VRiJyF3AXQO/eoVeRq1XFaQDiOo2otjvGaX3KHHVyGaXH3m1Y9Uop1Y4YY9i+fTuHDx/mjjvuAGDIkCE8+eSTdOzYMcrVqVAiGfIngV5VHvcM7KvqO8BNAMaYT0TEBaQBOVUbGWOeA54DGDNmTO3T14VQdO6stSHVu442fWV11/uO7wAgfsRUOPdVfQ6tlFLtRlZWFhkZGezevRuA0aNHc8kllwBowLdgkQz5zcBAEemLFe7zgYUXtDkGTAFeFJHLABdwhiZ07MylAJT6KideOHrWzck8a9S9Cdw//+tOnTXklVLqAiUlJaxcuZL3338fv99PXFwct99+Oz169Ih2aSoMEQt5Y4xXRO4H/oW19NsLxpjdIrII+NwYswr4EfC/IvJ/sAbhfctUnYausQo+ItFlnbF365UU3J25MTu4PdF8BsCH+z8GdGS9Ukqdt379elasWEFxcTEiwqRJk7j99ttJTk6++ItVixDRa/LGmLewbouruu+xKtt7gMil6pFfANaqRh1TnMHdPp/1OeLyPkkkHCir9hIdWa+UUpaDBw9SXFzMoEGDWLhwYf3HRKmoi/bAu8gq3MT5kK9q8ZpjgBXyHKjxtFJKtUtnz56lpKQkeK191qxZDBs2jLFjx+pMda1U2w5546mxq7iscqW5AftfAnRRGqVU+1ZeXs6aNWtYs2YNnTt35vHHH8fhcNChQwfGjRsX7fJUI7TdkM9eEnJ3bnFFcHvS0WcBXZRGKdU+GWP47LPPWLZsGXl5eQD06tWLiooKHI62Gw/tSdv9V/Tkhtz9+7eOVXv8276XgacE0Ln9jssAACAASURBVOvxSqn24+jRo7zyyiscOGBds+zduzcLFy5k0KBBUa5MNaW2G/I51hS2XxyZHtzl9Rn+8ZE1OU5fY4X95kDA61m8Uqq98Hg8LF68mIKCApKSkpg1axbXXHONTkXbBrXdkC+xJrlJSS6kqCQOu104W1TZVf+Q7y/VmutZvFKqLfN6vRhjcDqdOJ1OZs+ezfHjx5k5cybx8fHRLk9FSNsNeVcfcO9HYqxBdb36uIJPpZlzDGdfta56pZRqq3bs2EFmZibjx4/ntttuA2DCBO29bA/absgH2bHm2YHMjVZXvRc7oF31Sqm27fTp02RmZrJjh9WzuXXrVm699Vbtlm9H2nzIl7orJ9D79RtHAMiXDtXaaFe9UqotKS0t5c033+Tdd9/F5/MRFxfHzJkzmTJligZ8O9PmQ/5MYKkbsQkOm+D1G/6v7y+6II1Sqk06d+4cixYtorCwEBFh4sSJ3HHHHXTo0OHiL1ZtTpsP+fO69YzF67fO6q8zn/LrTtM15JVSbU5KSgo9evQgPT2dhQsX0qdPn2iXpKKo3YT8qYLy4HYSJbogjVKqTTh37hzLly9n+vTp9OjRAxHh3nvvJT4+XqeiVe0n5A9muwFrZL2Dyqlt9Xq8Uqo1qqio4O233+att96ioqKCsrIyHnjgAQASEhKiXJ1qKdpmyBsD7v3Vdm340pqy8ayk6vV4pVSrZYxhy5YtLF26lNxca2bPMWPGMHfu3ChXplqithnyvuIau97deQ6AoWYf3f8zAx67ormrUkqpRsnOzubvf/87e/fuBaBnz54sXLiQwYMHR7ky1VK1zZAP8JvKby8tyclXp2CCf0sUK1JKqYaz2WwcOHCAxMRE7rjjDiZOnIjdbo92WaoFa9Mhn5U/JLhtCwxAGYZ20yulWgev18vnn3/OuHHjsNlsdO7cmfvuu4/+/fuTmJgY7fJUK9A2Q95fBoDX5wSgS/dYzuKOZkVKKVUvu3fvJiMjg6ysLHw+X3Aa2pEjR0a5MtWatM2QL9kFwPGcXgC4XDY+2pcffPr+lx+ISllKKXUxOTk5LF26lK1btwLQuXNnkpOTo1yVaq3aZsj7Cqwvzj4A5OV6sAn4DSQYt94jr5RqcdxuN6tXr+add97B6/USGxvLjBkzmDZtGk6nM9rlqVaqbYZ8sbUYw/7j/QG4fGwyK7bnUlLuozcngXRA75FXSrUcH3/8MW+99RYAV199NbNmzSIlJSXKVanWrm2GvN2aCGL3kVEAFBV6o1mNUkqFVFxcHBxAN2nSJA4ePMjUqVPp379/lCtTbUXYyxGJSHwkC2lS7gMAdEy2BuCNHJOM8ejAO6VUy5CXl8fzzz/Pww8/TGFhIQAOh4O7775bA141qYuGvIhcLSJ7gL2BxyNF5E8Rr6yhfG44/TcARKzpa3td4gKftb24T7+olaaUat88Hg+rV6/mkUceYePGjZSXl3PgwIFol6XasHC66/8fcCOwCsAYs11EJka0qsbI+nPltrNTjae/8Fqz4emgO6VUczHGsG3bNjIzMzlz5gwAl19+OfPmzSM9PT3K1am2LKxr8saY4xesZuSrrW3Unfrf4GZ5hRPwhGymg+6UUs0lMzOTtWvXAtC9e3cWLFjA0KFDo1yVag/CCfnjInI1YETECfwA+DKyZTVC2WEA/L0XcTanMuBLJS5aFSml2rmxY8eyceNGbrvtNq6//nqdilY1m3BC/nvA74EewEngHeDeSBbVYMYEN7O90wE/ANtzay5Yo5RSkeDz+diwYQPHjx/nG9/4BgADBgzgd7/7HbGxsVGuTrU34YT8pcaYr1XdISITgI8jU1IjFG4Kbm7dkQJYyzB+/FVBlApSSrUnX375JRkZGZw4cQKAiRMn0qdPHwANeBUV4YT8H4HRYeyLPnfl4jMxrhgAhl2exPosK+QvN7s5HZXClFJt2dmzZ1m6dClbtlirXHbq1Il58+ZxySWXRLky1d7VGvIichVwNdBZRH5Y5alkoGVeUCq25nom+UoI9NB36RrDhx9b89aP929jZZRKU0q1TStXrmT16tV4vV5iYmKYPn06N954IzExMdEuTak6z+RjgMRAm6Qq+wuB2ZEsqsGynrW+xg2qtttmfPjFzg2DHKwsiUJdSqk2q7i4GK/Xy5VXXsns2bNJTU2NdklKBdUa8saY9cB6EXnRGHO0GWtqOEcqeM9Bp+nsXF4EWGPx/GJ1PGR0c4DOO6GUaoQjR45QXl7OpZdeCsBtt93GuHHjGDhwYJQrU6qmcK7Jl4rIU8BQwHV+pzFmcsSqaijvOetrh2tI7lAKwJfHK0/dPz6gq88ppRqmoKCAFStW8PHHH9O5c2eeeOIJnE4niYmJGvCqxQon5F8BlgIzsG6n+yZwJpJFNUhFTuW2LRawQv7D0/k1mupEOEqpcHm9Xt59911WrVpFWVkZdrud0aNH4/f7o12aUhcVTsh3Msb8VUR+UKULf3OkC6s3X5WL7fZEIA+A2Ip8kBj+w5fBP6NTmVKqldq+fTuZmZlkZ2cDMGLECObPn0/Xrl2jXJlS4Qkn5M9PG3dKRKYDWUDLHVkSW/2WFZunFGJiuG5ADP/UheiUUmHyeDy89NJL5OXl0bVrV+bPn8+IESOiXZZS9RJOyP9SRDoAP8K6Pz4Z+M+IVtWETpEOVJC+4FfwwqRol6OUasFKS0ux2Wy4XC6cTicLFiwgNzeXKVOm4HCEtdSHUi3KRX9rjTHne7kLgOshOONdi1bmrr6GzrPrfxqlSpRSLZ3f7+ejjz5ixYoVXHPNNcyZMweAMWPGRLkypRqnrslw7MBcrDnr3zbG7BKRGcAjQBxwefOU2DDbt1i30NmAvuY4nxzUkfVKqZq++uorMjIyOHrUulP4yJEj+P1+bDZblCtTqvHqOpP/K9AL+Az4g4hkAWOAh4wxbzRHcY2R3jWGnNMVFNn8JKasCO7XkfVKKYBz586xbNkyPvvsMwBSUlKYO3cu48aN44KltZVqteoK+THACGOMX0RcwGmgvzEmt3lKa7iSYi85pysA8Igfr+sYoGfxSinLmTNn+NnPfkZFRQVOp5Obb76Zm2++WReRUW1OXSFfYYzxAxhjykTkUGsIeIBjh8uC2yU2f3BOXj2LV0oBdO7cmcGDBxMbG8ucOXNIS0uLdklKRURdIT9YRHYEtgXoH3gsgDHGXPReEhG5CWstejvwvDHmtyHazAUeBwyw3RizsH7fQqj3tb6esXsw2uumVLt37Ngxli5dyrx58+jduzcA9913H06nM8qVKRVZdYX8ZY05cGDg3jPADcAJYLOIrDLG7KnSZiDwMDDBGJMnIukNfkNT3phylVJtUGFhIW+88Qbr16/HGMPKlSv5/ve/D6ABr9qFuhaoaeyiNOOAA8aYQwAikgncBuyp0uY/gGeMMXmB98ypcZRwFW+3vlacgoQGH0Up1QZ4vV7WrVvHypUrKS0txW63M3nyZGbOnBnt0pRqVpGc3aEHcLzK4xPA+AvaDAIQkY+xuvQfN8a8feGBROQu4C4g2NVWgwlMzBd/GSePVV6T791xcXDKPqVU23fs2DH+8pe/cOrUKQCGDh3KggUL6N69e5QrU6r5RXsKJwcwEJgE9AQ2iMhwY0y1VWWMMc8BzwGMGTPGhDxS/gbra2x3PB6rSYLfhkdH1ivVriQlJZGbm0t6ejrz589n5MiRekucarfCCnkRiQN6G2P21ePYJ7Husz+vZ2BfVSeAT40xHuCwiHyFFfr1XwDHGRgd6+hgDQ0ETjorgk/ryHql2ia328369euZNm0aNpuNlJQUfvzjH3PJJZfodXfV7l005EXkVuB3QAzQV0RGAYuMMRe7uLUZGCgifbHCfT5w4cj5N4AFwN9EJA2r+/5Q/b6FgJKd1teE4Q16uVKqdfH7/WzcuJEVK1ZQUFBAbGws119/PQADBgyIcnVKtQzhnMk/jjWI7gMAY8y2QHDXyRjjFZH7gX9hXW9/wRizW0QWAZ8bY1YFnpsmInsAH/Bgg+/FrzhtffUVN+jlSqnW4+DBgyxZsoTDhw8D0L9/f/r06RPdopRqgcJaatYYU3DBNa3Q18UvYIx5C3jrgn2PVdk2wA8D/zXO+TP55AvH9iml2oq8vDyWL1/OJ598AkDHjh2ZM2cO48eP17nmlQohnJDfLSILAXvgvvYHgI2RLasB7InWWXzCCHZuLYp2NUqpCNi2bRuffPIJDoeDG2+8kenTp+NyuaJdllItVjgh/33gp0A5sASri/2XkSyqQc5309uT2P+l1eMfa/STvVKtmTGGnJwcunTpAsDEiRPJzs5m8uTJpKc3fO4spdqLcEJ+sDHmp1hB3zJVVJlDR+zBzYMxZcRHoRylVOOdOHGCjIwMDh06xK9//WtSUlKw2+3Mnz8/2qUp1WqEE/L/LSJdgeXAUmPMrgjXVH/evOCmu7xyFSm3+DXklWpliouLeeONN1i3bh3GGBISEsjKyiIlJSXapSnV6lw05I0x1wdCfi7wFxFJxgr7ltNlXx6YWC9uEF5v5ZjAhNTnolSQUqq+fD4f69ev5/XXX6ekpAQRYfLkydx+++0kJiZGuzylWqWwJsMxxpwG/iAi64D/CzxGS7ouX2bNaof7q+CucvHjcn0J6Gx3SrUGf//73/noo48AuOyyy1iwYAE9e/aMclVKtW7hTIZzGTAPmAXkAkuBH0W4rvrJX2d9Tbs95NM6251SLZMxJjjl7OTJk9m3bx9z585l9OjROhWtUk0gnDP5F7CC/UZjTFaE62kYZ2CUrT05uMsmZbU0VkpFW3l5OatXr+b06dPce++9APTp04ff/OY3er+7Uk0onGvyVzVHIU0icUSVBz5Au+qVakmMMWzatInly5eTl2cNmD1+/Di9elnLXGjAK9W0ag15EVlmjJkrIjupPsOdYE1WN6KWl0ZVXu75hWWtrj7tqleqZTh8+DBLlizh4MGDgHXmvnDhwmDAK6WaXl1n8j8IfJ3RHIU0lYJ8K+Sdfh2Nq1RLYIzh5Zdf5oMPPgAgOTmZWbNmMWHCBD1zVyrCag15Y8ypwOa9xpifVH1ORP4L+EnNV0VfUZnVTZ8fvzPKlSilAEQEl8uF3W5n2rRpzJgxg7i4uGiXpVS7EM7AuxuoGeg3h9inlFIYY9i+fTs2m40RI6yrerfeeivXXXddcHpapVTzqOua/D3AvUA/EdlR5akk4ONIF9ZQX+0vCW7roDulmldWVhaZmZns2rWL1NRUfvWrXxEbG0tcXJyevSsVBXWdyS8B1gC/AR6qsr/IGHMuolU1QmGxF4BYT2cddKdUMyktLWXlypW8//77+Hw+4uLimDZtGna7/eIvVkpFTF0hb4wxR0TkvgufEJHUFhX05yfDAXyBaW3PJX2GdaVBKRUpfr+fDRs28Nprr1FcXIyIMGnSJG6//XaSk5MvfgClVERd7Ex+BrAF6xa6qtNPGaBfBOuqH3tgGRp/GVkHrElwxIQ1Y69SqhF8Ph9vvfUWxcXFDBo0iIULF9K7d+9ol6WUCqhrdP2MwNe+zVdOA0ng20i+ioQODgpyvHgc+dGtSak2Kjc3F5fLRUJCAk6nk2984xuUlpYyduxYnYpWqRYmnLnrJwDbjDElIvJ1YDSw2BhzLOLV1Zuw/3QpCdgoc+ZcvLlSKmzl5eWsWbOGNWvWMHHiRL72ta8BMGzYsChXppSqTTh92n8GRorISKyFaZ4HXgaui2RhDZVQZE2uIegZhVJNwRjD5s2bWbZsGefOWUNxSkpKqi0uo5RqmcIJea8xxojIbcDTxpi/ish3Il1YQ/mcBrtHKHe0nHGBSrVWR48eJSMjg6++spZx7t27NwsXLmTQoEFRrkwpFY5wQr5IRB4G/g24VkRsgDOyZTVMuddPhdcQh+BxFES7HKVatezsbBYtWoQxhsTERGbNmsW1116rU9Eq1YqEE/LzgIXAt40xp0WkN/BUZMtqmAOn3cSZDtEuQ6lWy+/3B0O8S5cujBs3jg4dOjBz5kzi4+OjXJ1Sqr7CWWr2tIi8AowVkRnAZ8aYlyJfWj0UbgLgwDFvcJdfKqJVjVKt0s6dO8nMzOTb3/42/fv3B+Cuu+7S6+5KtWLhjK6fi3Xm/gHWvfJ/FJEHjTHLI1xbeMqOBzf/+lYpXQPbVw++PDr1KNXKZGdnk5GRwY4d1uzVa9euDYa8BrxSrVs43fU/BcYaY3IARKQz8C7QMkLeW3k/vLvUWpe6MG4vmTqlrVJ1crvdvPnmm6xduxafz4fL5WLmzJlMnTo12qUppZpIOCFvOx/wAblAyxt5kzAMe6Ashy8hysUo1bLt37+fZ555hsLCQkSEa665hlmzZtGhg45pUaotCSfk3xaRfwEZgcfzgLciV1LDuXxWyOcmbQZuiW4xSrVgXbp0wePx0L9/fxYuXEjfvi1/YkulVP2FM/DuQRG5E7gmsOs5Y8zrkS2rYeICIW/3x0a5EqValry8PNauXcudd96Jw+EgOTmZRx99lK5du+p1d6XasLrWkx8I/A7oD+wEfmyMOdlchYXNVI6oL/X7SMJOietoFAtSquXweDy8/fbbrF69moqKCjp27Mi0adMA6NatW5SrU0pFWl1n8i8ALwEbgFuBPwJ3NkdR9VL0OQCeiuLgLp/NHa1qlGoRjDF88cUXLF26lLNnzwJwxRVXcPnleteJUu1JXSGfZIz538D2PhH5ojkKqrfACnRnS1reWEClouHEiRMsWbKEvXv3AtCjRw8WLlzIZZddFuXKlFLNra6Qd4nI5VSuIx9X9bExpmWEfv77AJz2jYxyIUq1DIcPH2bv3r0kJCRwxx13cN1112G326NdllIqCuoK+VPA/1R5fLrKYwNMjlRR9eLoaH31FeF0ngBfv+jWo1Qz8/l8HD16lH79rN/9CRMmUFRUxMSJE0lMTIxydUqpaKo15I0x1zdnIQ1WtAWAd44Nx2YrAWBY9yHRrEipZrNnzx4yMjLIycnhV7/6FWlpadhsNm65RW8hVUqFd598y+YrAiAtwQR33TvlnmhVo1SzyMnJYdmyZXzxhXXVLC0tjYKCAtLS0qJcmVKqJWn9IW9PAmB/ydAoF6JU5JWVlbF69Wr+9a9/4fV6iY2NZcaMGUybNg2ns0WuAK2UiqLWH/IBZV47HUuHR7sMpSLqpZdeYtMma9XFq666itmzZ5OSkhLlqpRSLVU4q9AJ8DWgnzFmUWA9+a7GmM8iXl09HC/4Bx1lEXYTiytOb6dTbYfX68XhsP5XnT59OmfPnmXevHnBleKUUqo24ZzJ/wnwY42mXwQUASuAsRGsq95iYo7gFw92E0vX7jqtrWr98vPzWbFiBfn5+fzwhz9EROjRowePPPJItEtTSrUS4YT8eGPMaBHZCmCMyRORmAjX1SBOv94upFo/j8fD2rVrefPNNykvL8fhcJCVlUWPHj2iXZpSqpUJJ+Q9ImLHujf+/Hry/ohW1QC20kuC2w6nLrihWh9jDNu2bWPp0qXk5FirO48aNYp58+bRpUuXKFenlGqNwgn5PwCvA+ki8itgNvBoOAcXkZuA3wN24HljzG9raTcLWA6MNcZ8Hs6xL2Qr6x7cdrl0di/VuhhjePrpp9m6dStgLR6zYMEChg0bFuXKlFKtWThLzb4iIluAKVhT2t5ujPnyYq8LnP0/A9wAnAA2i8gqY8yeC9olAT8APm1A/UE2d28ALrWvB3SUvWpdRIRevXqxb98+brvtNq6//vrgYDullGqocEbX9wZKgTer7jPGHLvIS8cBB4wxhwKvyQRuA/Zc0O4J4L+AB+tRd806/dYwgSLXwMYcRqlm4ff72bBhA3FxcYwfPx6Am2++mSlTppCUlBTl6pRSbUU4pwqrsa7HC+AC+gL7gIvNPtMDOF7l8QlgfNUGIjIa6GWMWS0ijQp5xBomMH6mLqWpWrZ9+/axZMkSjh8/TnJyMiNHjsTlchEbG0tsrN4ZopRqOuF011fr+w4E872NfWMRsWEtePOtMNreBdwF0Lt375Bt7AVXnG/b2NKUioizZ8+ybNkyPv/cGnbSqVMn5s2bp8GulIqYel/0M8Z8ISLjL96Sk0CvKo97BvadlwQMAz4IBHNXYJWIzLxw8J0x5jngOYAxY8YYQjCOfKAHxh/yaaWixuPxsHr1atasWYPH4yEmJoZbbrmFm266iZiYFnk3qlKqjQjnmvwPqzy0AaOBrDCOvRkYKCJ9scJ9PrDw/JPGmAIguJqGiHwA/Li+o+v9xlB1fru+A+Pr83KlIk5E2Lx5Mx6Ph/HjxzNnzhxSU1OjXZZSqh0I50y+6iggL9Y1+hUXe5Exxisi9wP/wrqF7gVjzG4RWQR8boxZ1ZCCL+Su8BNnBEexLlCjWo6jR4+SkpJCcnIyDoeDb37zmwAMGjQoypUppdqTOkM+cBtckjHmxw05uDHmLeCtC/Y9VkvbSQ15D3eFj+KCyqsCqZ10JS4VPYWFhbz22mt8+OGHXHvttXzrW98CNNyVUtFRa8iLiCNwNj6hOQuqLxHBb6zBdg7KSO+qg5hU8/N6vbz33nusWrUKt9uN3W4nPj4eY4wOBlVKRU1dZ/KfYV1/3yYiq4BXgZLzTxpjXotwbfWWLDnRLkG1Qzt27CAzM5PTp08DMHz4cObPn0+3bt2iXJlSqr0L55q8C8jFWoXu/P3yBmhxIa9Uc8vKymLx4sUAdOnShQULFjBixIgoV6WUUpa6Qj49MLJ+F5Xhfl6LuU/tTNFJOqAzhKnmU1FREbz1rXv37kyePJm0tDSmTp2qU9EqpVqUuv4i2YFEqof7eS0m5H2mFDTkVTPw+/189NFHvPbaa9x9991cdtllAHz961+PcmVKKRVaXSF/yhizqNkqUaoF279/P0uWLOHo0aMAfPrpp8GQV0qplqqukG/5Q4K9RQxNLuNUXrQLUW3VuXPnePXVV/n0U2uRxJSUFObMmRNcVEYppVqyukJ+SrNV0VA5r0S7AtWG7dq1i6effpqKigqcTic33XQTt9xyi841r5RqNWoNeWPMueYspCGyjm2iO3CuTOf/Vk2vb9++xMTEMGLECObOnUtaWtrFX6SUUi1Iqx4KXFpeBgKbjlu3LFWg89arhjt+/Dhvv/023/rWt3A6nSQkJPDLX/6S5OTkaJemlFIN0qpDfoC8AcBXxdaZvIOyaJajWqmioiJef/111q9fjzGGnj17cvPNNwNowCulWrXWG/LGG9zMz70cgG62vcDNUSpItTZer5d169axcuVKSktLsdlsTJkyhYkTJ0a7NKWUahKtOOR9wc1THg9d0e56Fb59+/bx8ssvk5VlrZo8dOhQ5s+fT48ePaJcmVJKNZ3WG/LlJwHw+iGxrD8Al9i+iGZFqhUpKCggKyuL9PR05s2bx6hRo3QhGaVUm9N6Q77CWgzEYYP4sksAkK66prwKze12s3///uC88mPHjsXj8TBu3DicTl2eWCnVNrXekA/Ymh+HLfBtjPmP70S5GtXS+P1+PvnkE5YvX05xcTGLFi2iW7duiAgTJrToVZSVUqrRWn3Il5VXjn7u3TcuipWolubgwYMsWbKEw4cPA9Z97z6f7yKvahyPx8OJEycoK9M7PZRS9eNyuejZs2eT9i62+pAvKa2coMTptEWxEtVS5Ofns3z5cjZu3AhAhw4dmD17NldddRU2W2R/R06cOEFSUhJ9+vTRa/xKqbAZY8jNzeXEiRP07du3yY7b6kP+vM5yEBge7TJUC5CZmclnn32Gw+Fg2rRpTJ8+nbi45unlKSsr04BXStWbiNCpUyfOnDnTpMdtMyGv2i9jDG63m/h46xbKO+64A5/Px5w5c0hPT2/2ejTglVINEYm/Ha025L1+03qLV03m5MmTZGZmUl5ezsMPP4yI0KVLF+67775ol6aUUlHXai9il1X4o12CiqKSkhJeeeUVfv7zn7N7926ysrLIycmJdlktgt1uZ9SoUQwbNoxbb72V/Pz84HO7d+9m8uTJXHrppQwcOJAnnngCY0zw+TVr1jBmzBiGDBnC5Zdfzo9+9KOQ7xFuu0jp06cPZ8+erbYvOzubGTNmMHLkSIYMGcItt9zCzp07GTVqFKNGjSI1NZW+ffsyatQopk6dypEjRxARHn300eAxzp49i9Pp5P7776/xnqGOD3DkyBGWLFnSpN9LfeXn5/OnP/0p+DgrK4vZs2fX6xj/8z//w+DBgxk+fDgjR47khz/8IR6PJ1jj8OHDGT58OEOGDOHRRx+tdXCp2+3muuuuqzbIdfHixbhcLgoKCoL7XnzxxRo/50mTJvH5558DUFxczN13303//v254oormDRpUnDJ54YyxvDAAw8wYMAARowYwRdfhJ5bZenSpYwYMYKhQ4fyk5/8pFrNnTt3Dv5OPf/888HnbrrpJjp27MiMGTOqHWv+/Pns37+/UXU3RqsN+aNndfRye+Tz+Vi3bh0PP/ww7733HsYYrr/+en7zm9/QpUuXaJfXIsTFxbFt2zZ27dpFamoqzzzzDGD98Z05cyYPPfQQ+/btY/v27WzcuDEYDrt27eL+++/nH//4B3v27OHzzz9nwIABNY4fbrvaeL3eizdqgMcee4wbbriB7du3s2fPHn77298yfPhwtm3bxrZt25g5cyZPPfUU27Zt49133wWsOy5Wr14dPMarr77K0KGh59sIdXxofMg3hQtDvnv37ixfvjzs1z/77LO88847bNq0iZ07d7J582bS09Nxu93BNuvWrWPnzp189tlnHDp0iLvvvjvksV544QXuvPNO7HZ7cF9GRgZjx47ltddeC7um7373u6SmprJ//362bNnC3/72t0Z/GFqzZg379+9n//79PPfcc9xzzz012uTm5vLggw/y3nvvsXv3bk6fPs17770XfH7evHnB36nvfve7wf0PPvggL7/8co3j3XPPPTz55JONqrsxtMdbtRrGGP7rv/6LAwcOYgB5RQAAIABJREFUADB48GAWLFhAr169olxZaAe+3Tkixx3wQvgDc6666ip27NgBwJIlS5gwYQLTpk0DID4+nqeffppJkyZx33338eT/b+/Mw6qq2j58L3BCCUNNM3EWURQ4gDiUKKg4DyEmDjmVVqbZaFnOn1qWmulbZmY58oKoKU6J5pBmlgOC4jyhguSEIYggw/r+OJz9chgPKDK47us6F+fsvfZaz157s5+9puf31VdMnDiRJk2aAPoegewegrmlGz58OD179tRakZaWlsTHx7N3714mT56MtbU1Z86coW/fvtSuXVsbVpk2bRqWlpZ89NFHzJkzh8DAQJKSkvD29mb69OkmnWt0dLR2boAW+Cg3KlasSNOmTTly5AgtWrRgzZo19O/fXwt3bEr+EyZM4PTp0+h0OoYNG8bo0aMZPXo0R44coUyZMnz99dd4enqSmprKJ598wvbt2zEzM2PUqFG88847APznP/9h8+bNJCcns3btWpo0acKhQ4d49913SUxMxMLCgmXLlmFnZ8fJkycZMWIEDx8+JC0tjfXr1zN58mQuXryITqfDy8uLMWPG0LNnT8LDw3Mt18CsWbPYt28fzz77LADlypVjwoQJ2daZpaUlixcvpnbt2sTExFClShWj/X5+fkYvPRcvXiQ+Pp5FixYxa9YsRowYked1uXjxIn///Td+fn7aipj69es/8qzzoKAghg4dihCC1q1b8++//xIdHU3NmjW1NJcuXcLW1pbnntP//3bq1In169fTsWPHXPPu2LEje/fuzbLd3d2d4cOHk5KSQpkyT97lltiWfLm0aAAuXWtfxJYonhRCCJycnKhatSpvv/0248ePL7YOvjiQmprKrl276N27N6Dvqnd1dTVK07BhQ+Lj47l37x7h4eFZ9meHqekyExISwoIFCzh37hy+vr4EBgZq+wIDA/H19WXHjh2cP3+eQ4cOERoaytGjR9m3b59J+Y8ZM4bXX38dT09PZs2ala2jzo4BAwYQEBDAtWvXMDc354UXXshX/rNnz8bd3Z3Q0FDef/99vvvuO4QQnDhxAn9/f4YNG0ZiYiJLliwhIiKC0NBQjh8/zuDBg7W8q1WrRkhICKNHj2bu3LmA/iV2//79HDt2jP/7v//js88+A/St7nfffZfQ0FCOHDmCjY0Ns2fPpmHDhoSGhjJnzhwju3MrF+DevXvEx8fny4FaWVlRv379LN3QDx8+5NKlS9SrV0/bFhAQwIABA3B3d+fs2bPcuHEjz/xPnjyJTqcz6g3ICUNY6syflStXZkkbFRVl9MywsbEhKirKKE2jRo04e/YsERERpKSksHHjRq5du6btX79+PY6OjvTr189oe06YmZnRqFEjwsLC8kxbGJTYlnz5FH3lyiT9OvlbsmFRmqMoBJKSkti2bRtVqlShfXv9y1znzp3x8vKiXLlyRWxd3uSnxf04efDgATqdjqioKJo2bYqXl1eR2JGZli1bao7E2dmZmzdvcv36dW7duoW1tTW1a9dmwYIF7NixA2dnvbJkfHw858+fN0kZsEuXLly6dInt27fz66+/4uzsTHh4uNYiy4muXbsyefJkatSoga+vb77zz8wff/yhtZSbNGlC3bp1OXfuHL/99htvvfWW1prL2ALu27cvAK6urlqXdmxsLMOGDeP8+fMIIbTx8TZt2jBr1iwiIyPp27cvtra2uZ5fbuVmR3BwMJ988gn//vsv//3vf3nxxRezTZdxLoeB27dva70BBvz9/dmwYQNmZmb4+Piwdu1axo4dm+NM8vzOMF+zZk2+0ueFtbU133//Pb6+vpiZmfHiiy9y8eJFAHr16sXAgQMpX748P/zwA8OGDWP37t155lm9enWuX79eoJfjR6XEtuSfTdKPqd3U3/d0LPufIrRG8TiRUvLXX3/x2WefsXnzZtatW0dSUhIAZcuWLREOvigxjMlfuXIFKaU2Jm9vb8/Ro0eN0l66dAlLS0usrKxo1qxZlv3ZkVu6MmXKkJamnxSblpbGw4cPtX2VKlUySvvKK6+wbt061qxZozlXKSWffvqpNuZ54cIFXn/d9HDVVapUYdCgQaxatQo3NzeTegHKlSuHq6sr8+bNy3OyWkHyN4Xy5csD+qEPw5yFyZMn4+npSXh4OJs3b9Ymug0aNIhNmzZhYWFB9+7dTXIyuWFlZYWlpaUWGbJLly6EhobSvHlzo+uXkbi4OCIiImjcuLHRdgsLC6MJeSdOnOD8+fN4eXlRr149AgIC8Pf3B6Bq1arcvXvX6PiYmBiqVatGs2bNCAsLMylCZX5a8rVq1TJqfUdGRmarPNmrVy/+/vtvDh48iJ2dnXaeVatW1a7VyJEjTfp/AbQhl6KgxDr5h+b6MZQUqX/rq1Ar97dZRckgIiKCL774giVLlnD37l3q1q3LuHHjtH8shelUrFiRhQsXMm/ePFJSUhg8eDB//PGHNunswYMHjBs3jo8//hjQTxz6/PPPOXfuHKB30osXL86Sb27p6tWrpz34Nm3apLU+s8PX15eAgADWrVvHK6+8AugdzM8//0x8fDyg7141ddXE7t27SUhIAPRO6OLFi9SpU8ekYz/88EO+/PLLXFu5OeX/zDPPEBcXp6Vzd3fHz88PgHPnznH16lXs7Ozw8vLihx9+0Jx4TExMrjbFxsZqDmj58uXa9kuXLtGgQQPGjRtHnz59OH78eBYbMmJKuZ9++imjR4/WVmJIKXOcPR8fH8/bb7/Nyy+/jLW1tdE+a2trUlNTtWP9/f2ZNm0aERERREREcP36da5fv86VK1dwc3PjwIED/POPXmzsyJEjJCUlUbt2bRo2bEiLFi2YOnWq1mMQERFhNEnSwJo1a7SXwoyfoUOHZknbu3dvVq5cqTUkKleubDQeb8Bwz929e5dFixZpE+yio6O1NJs2baJp06bZ1lFmzp07R/PmzU1K+7gpsd31BmIe6sdsKnd4rYgtUTwKiYmJ+Pv788cffyClxMrKCh8fH1566aVCD0VbmnF2dsbR0RF/f3+GDBlCUFAQ77zzDmPGjCE1NZUhQ4Zoy5gcHR355ptvGDhwIAkJCQghsiwHyivdqFGj6NOnD05OTnTt2jVL6z0jzZo1Iy4ujlq1amkP2s6dO3P69GnatGkD6Cd5rV69OtugRo6Ojtq90b9/f2rWrMnYsWO13oSRI0fi5uZmUj01a9Ysx1n1Bo4ePZpt/snJyZibm+Pk5MTw4cN5++23GT16NA4ODpQpU4bly5dTvnx5Ro4cyblz53B0dKRs2bKMGjUq26V6Bj7++GOGDRvGzJkz6dGjh7Y9MDCQVatWUbZsWZ5//nk+++wzqlSpwksvvUTz5s3p1q2bUZwIU8odPXo09+/fp1WrVpQvXx5LS0teeuklbdgEwNPTEyklaWlpeHt7M3ny5Gzt7ty5M3/88QedOnUiICCAbdu2Ge339vYmICCATz75hAULFtC9e3fS0tKwtLTE399fu6ZLly7lww8/pFGjRlhYWFCtWrUs8w3yS/fu3dm2bRuNGjWiYsWKLFu2TNun0+kIDQ0F4N1339XG0KdMmaK15BcuXMimTZsoU6YMVapUMXr5cnd358yZM8THx2NjY8NPP/1Ely5duHHjBhYWFjz//POPZHtBEdmNqxRnWrRoIY8cOcLtw4Oo9uAX3tnwfyRFd2PQay/g0blqUZunKCBpaWlMnz6d69ev4+XlRc+ePbUIdiWJ06dPm/x2r1CURkJCQpg/f362y8meRubPn4+VlZXJw07ZPUOEEEellC0KUn6Jb8krSi5hYWHUrl2bKlWqYGZmxmuvvUb58uWL7I1XoVA8Oi4uLtqSQVNmx5d2nn32WYYMGVJk5Zd4J58U3a2oTVDkk+joaAICAjhx4gStW7fmjTfeAKBu3bpFbJlCoXgcvPaaGj41YEpcgMKkZDp5Kan24Bdu3/vfetZKluqNsbiTkJDApk2b2LVrF6mpqVhYWFCvXj2klErURaFQKAqBkunkU/WzSOMf/G89pnNLq6KyRpEHaWlp7N+/n19++YW4uDiEELRr146+fftiZaWum0KhUBQWJdPJp3MrTh+r/HmzM5Qpo7TkiyvR0dHaspXGjRszcOBA1TWvUCgUT4ASvTbpdrw+ktWdNOUwihv37t3TvteqVYtevXrx1ltv8cknnygHX8goFbrCUaFbvnw5QggtzgDAxo0bEUJoYjAjR47k1KlTudq+ePHibAO1TJs2jVq1aqHT6WjSpAmjR4/WAgvlxMaNG/Msb+/evdkuhTTQr18/Ll26pP0ODQ1FCMH27du1bREREVnWeU+bNk0LwQswd+5cmjRpgk6nw83NLdtzzC8rVqzA1tYWW1tbVqxYkW2asLAw2rRpg4ODA7169dKePREREVhYWGjX/6233tKO8fDwwM7OTttnWBf/7bff8vPPPz+y3cUKKWWJ+ri6ukqZHCvl7+Xl9v8MlqN8j8v/DPpCKooHiYmJMigoSL755psyLCysqM154pw6daqoTZCVKlXSvg8dOlTOnDlTSillQkKCbNCggQwODpZSSnn//n3ZtWtX+e2330oppTxx4oRs0KCBPH36tJRSypSUFLlo0aIs+ZuaLieSk5MLdmIZqFu3rrx165bRtjfeeEN+88032u/M99+wYcPk2rVrtd+XL1+W9evXlzqdTtu2aNEi6eTkJMeMGZOlzGXLlkkHBwf5+uuva9v69+8vnZycjPItKFOnTpVz5syRUkqZmpoqX3rpJbl79+5cj8l8TtmxZ88e2aNHj2z3hYeHy5dfftlo28cffyzbtm0rhw4dqm27fPmybNasWY72fv/997Jz584yNjZWSillbGysXL58ea525cWdO3dk/fr15Z07d2RMTIysX7++jImJyZKuRYsWcu/evVJKKX/66Sc5adKkHG020L59e3n48OEs2+/fv290PxQF2T1DgCOygD6zRLfkDZiJvEMfKgoXKSWHDh1i0qRJbNy4kYcPH3LmzJmiNuupp02bNpoAR04qdAbJ1MelQpdR4tTS0hLQtybd3d3p3bs39vb2TJgwQQu3C8atwjlz5uDm5oajoyNTp041+Vyjo6OxsbHRfudXhQ7QVOhywt3dnUOHDpGcnEx8fDwXLlxAp9Np+zPqoVtaWjJx4kScnJxo3bq1JsySuQWcHQ8fPiQxMVGLKPfjjz/i5uaGk5MTPj4+JCQk8Oeff7Jp0ybGjx+PTqfj4sWLXLhwgU6dOuHk5ISLi4sWcz0+Pp5+/frRpEkTBg8erPXe+Pn50adPH61cKSVr165l+fLl7Ny5M8eod5n5/PPP+f7777U5NlZWVgwbNsykY3MiODgYLy8vqlSpgrW1NV5eXka9CwbOnTunaRt4eXmxfv36ApdZsWJF6tWrx6FDhwqcR3GjRI/JK4oHV69exd/fn7NnzwJQu3ZtBg0ahJ2dXRFbVrRUf+P3Qsn35hLTlBcNKnSGIBymqNCZ0u1uarrMhISEEB4eTv369Tl27BjvvfeeFpktMDCQ4OBgIxU6KSW9e/dm3759JgnUjBkzBl9fX7799ls6derEiBEjclSUy4hBha5GjRqaCl1OCnZCCDp16kRwcDCxsbH07t1bi/memfv379O6dWtmzZrFxx9/zI8//mg0NJAd8+fPZ/Xq1Vy5coVu3bppLxB9+/Zl1KhRAEyaNImffvqJd955h969extJ+7Zq1YoJEybg7e1NYmIiaWlpXLt2jWPHjnHy5EleeOEFXnrpJQ4cOEDbtm05cOAAAwcO1Mr/888/qV+/Pg0bNsTDw4OtW7fi4+OTq8337t0jLi6OBg0a5JoO9C9whpC/GWnXrh0LFy402maKYhzooxUGBQXx8ssvs3btWqPY9JcvX8bZ2RkrKytmzpyJu7u7tm/EiBGYm5vj4+PDpEmTtBU+LVq0YP/+/bRs2TLP8ykJlMyW/AN9zOwyomRF6yuNHD16lOnTp3P27FksLS0ZOnQoU6dOfeodfFFiUKF7/vnnuXHjRrFXoQsLC9NU6Hbs2KGp0Lm4uHDmzJkscqY5YVCJGzVqFGfOnMHZ2Zlbt/JWAuzatSs7d+4kICAgVxU6A4aXgoCAACMHmZly5cppY+Gurq5ERETkmff7779PaGgoN2/e5P79+wQEBAD6Fyt3d3ccHBzw8/Pj5MmTWY6Ni4sjKioKb29vACpUqKBFjWzZsiU2NjaYmZmh0+k0W6Kjo41U+vz9/RkwYIB2ngYxmcelGDd+/Phs48xndvD54eeff2bRokW4uroSFxenCVjVrFmTq1evcuzYMb7++msGDRqkjdf7+flx4sQJ9u/fz/79+42i8xkU40oLJbMln6JfQnfr33pFa4cCe3t7rK2tcXFxoU+fPrnGKn/aMLXF/bgxqNAlJCTQpUsXvvvuO8aNG4e9vX0W1bTsVOicnJxyzT+3dAVRofvnn3+yqNC9+eabBTp3g0rcoEGD6NmzJ/v27cuzJZpRhe7UqVNs2rQp1/QtW7bkxIkTVKxYMYsKW0bKli2rOcGM6nKmULZsWbp27cq+ffsYMGAAw4cPZ+PGjTg5ObF8+XL27t1rcl6AkcBTRlsyqsalpqayfv16goKCmDVrFlJK7ty5Q1xcXI6KcfXr19dU7AzCObmRn5Z8rVq1jM4zMjISDw+PLMc2adKEHTt2APque4OITfny5bXzdnV1pWHDhpw7d44WLVpowj/PPPMMgwYN4tChQ5qgTVEqxhUGJbMln87uE/o3zvsyd31kxeMjPDycuXPnag8GCwsLZs2axaBBg5SDL2YoFbrHq0KXkdmzZ/P555+blLYgSCk5cOAADRs2BPTnU7NmTZKTk42cZEb1uWeeeQYbGxs2btwIQFJSklYfOdG0aVMuXLgAwK5du3B0dOTatWtERERw5coVfHx82LBhA5aWltSsWVOTtY2JiWH79u20bdsW0KvYjRkzRmspx8fHZzu7Pj8t+S5durBjxw7u3r3L3bt32bFjB126dMmSznB/pKWlMXPmTG0W/a1btzSp2kuXLnH+/HkaNGhASkqKtjIjOTmZLVu2GK0cKErFuMKgxDr5NPm/bqKm5ruK0JKngxs3brBw4UK+/vprTp06xZ49e7R9Sga2+JJRhc7CwoKgoCBmzpyJnZ0dDg4OuLm5ZatC17RpU5o3b260tMpAbulGjRrF77//jpOTEwcPHiyQCt2gQYO0JVH9+vXLUULV0dERGxsbbGxs+OCDDzh69CgtWrTA0dGRNm3a5FuFLj8Txbp164anp6fJ6U1l/vz52vLH1NRU3n77bQBmzJhBq1ateOmll7QJj6DvUp8zZw7Ozs5cvHiRVatWsXDhQhwdHXnxxRc1Gdec6NGjh9Za9vf317r6Dfj4+Ghd9itXrmTGjBnodDo6dOjA1KlTtZeQ0aNH4+npiZubG82bN8fd3f2R1SOrVKnC5MmTcXNzw83NjSlTpmgvYSNHjtQmOPr7+9O4cWOaNGnCCy+8oIWR3bdvH46Ojuh0Ovr168fixYupUqUKSUlJdOnSRdtXq1Ytbb4DwIEDB4rNENfjoFBV6IQQXYEFgDmwVEo5O9P+D4CRQApwC3hNSnkltzxbtGghj+ycQ3Job8b88CcAk1rOps4HWbuAFI/OgwcP2LJlCzt27CA1NZUKFSrQq1cvOnXqRNmyZYvavGKHUqFTlCQePHiAp6cnBw4cUGIyoI3fF6WCXolRoRNCmAPfAV5AJHBYCLFJSpkxcsMxoIWUMkEIMRr4Csh75kvGckhRDr6QCAkJYdWqVcTGxgLQtm1bfHx8qFy5chFbplAoHgcWFhZMnz6dqKgok4c2SjO3b99mxowZRW3GY6UwJ961BC5IKS8BCCECgD6A5uSllHsypP8LeNWUjB88uENacumZGFFcMTMzIzY2loYNGzJo0CBtZrRCoSg9ZDfO/bRSmrrpDRSmk68FXMvwOxJolUv614FfTcn46vEPuHzmZQBkCV0gUBy5e/cup0+f5sUXXwTAycmJDz74AHt7+0ceX1MoFArFk6dYeEghxKtACyDbNUdCiDeANwDq1KnDtfgHlDHXz9qtLK4DSpzmUUhOTiY4OJitW7fy8OFD6tSpg42NDUKIUjXLVKFQKJ42CtPJRwG1M/y2Sd9mhBCiEzARaC+lTMouIynlEmAJ6Cfepcmr3E+XmW1kfhBQ3U0FQUpJSEgIa9as0ZaUuLi4UKFChSK2TKFQKBSPg8J08ocBWyFEffTOfQAwKGMCIYQz8APQVUpp2mLYdC7f1LcwH8qKj8XYp43IyEj8/f05ffo0oA88MXDgQOzt7YvYMoVCoVA8LgptoFVKmQKMBYKB00CglPKkEOL/hBC905PNASyBtUKIUCFE7qGmMnAmUh9XuIrZtTxSKrJj69atnD59mkqVKjF48GCmTZumHHwpQUnNFo7UbHb5g17S9L///W+BzyWjqE1+ySwIlNf2ouSbb74xCpCTkpLCc889x4QJE4zSZb62maVyC+PeO3r0KA4ODjRq1Ihx48aR3dLyu3fv4u3tjaOjIy1btiQ8PNxof2pqKs7Ozka27tq1CxcXF3Q6HW3bttUCDz1RSduCytcV1cfV1VUGBz4nJ725WY7yPS53Du2Sp3SfQi8HmlGm8fbt23L16tUyLi6uCK0qfSip2bwpqVKzOeWfm5SrKeQke2oKOUnNmiJB+zhIS0uTqampeaZLTk6WDg4ORtd+27Zt8sUXX5QNGjSQaWlp2vbM1zZj/T7qvZcTbm5u8uDBgzItLU127dpVbtu2LUuajz76SE6bNk1KKeXp06dlhw4djPbPmzdPDhw40OhesLW11Z4J3333nRw2bJiUMndJ26deajYlTeJZNZEb/9YF4BmRr17+p5LTp08zbdo0Fi5cqMUVr1q1KoMHD9akQBWlEyU1mzv5kZrNKf8JEyawf/9+dDod8+fPJyIiAnd3d1xcXHBxceHPP//Ujvnyyy9xcHDAyckpSws2LS2N4cOHM2nSJFJTUxk/frxWDz/88AOgb5SNHTsWOzs7OnXqZHLIX9CHmu3YsSMuLi44ODgQFBQEwJQpU/jmm2+0dBMnTmTBggVA9tciIiICOzs7hg4dSvPmzbl27RrDhw+nefPmODg4MH/+/Cxl7969GxcXF8qU+d8Isb+/P++++y516tTh4MGDJp2DqfdofoiOjubevXu0bt0aIQRDhw7VQgNn5NSpU3To0AHQx8uPiIjQ5IMjIyPZunUrI0eONDpGCKGF+o2NjdVUEZ+kpG2xmF2fL2Qqxy57aD8txe2c0z7l3Lp1izVr1hASEgJAtWrViImJoVq1akVs2dOB0xTXvBMVgLD/O2pSOiU1+3ilZnPKf/bs2cydO5ctW7YAkJCQwM6dO6lQoQLnz59n4MCBHDlyhF9//ZWgoCD+/vtvKlasSExMjJa3QVugefPmTJw4kSVLllC5cmUOHz5MUlKS9nJ27Ngxzp49y6lTp7hx4wb29va89tprJtV/hQoV2LBhA1ZWVty+fZvWrVvTu3dvXnvtNfr27ct7771HWloaAQEBHDp0KMdrUadOHc6fP8+KFSto3bo1R48eJSoqSuu+zjg8ZODAgQNG915iYiK//fYbP/zwA//++y/+/v7a0t3cMPXe27NnD++//36W7RUrVjR66QK9PkLGl7ecJG2dnJz45ZdfcHd359ChQ1y5coXIyEhq1KjBe++9x1dffZUlBPPSpUvp3r07FhYWWFlZ8ddff2n7npSkbYlz8kKm8s9dfSu+LA8oLx4UsUXFj8TERLZu3UpwcDApKSmUL1+eHj160KVLFxWK9inAIDUbFRVF06ZNi02Aj5ykZm/duqVJzS5YsECTmgV96/P8+fMmOXmD1Oz27dv59ddfcXZ2Jjw83EhKNTu6du3K5MmTqVGjRq5Ssznln5nk5GTGjh1LaGgo5ubmmpDPb7/9xogRIzT514xiOG+++Sb9+/dn4sSJAOzYsYPjx49rvSKxsbGcP3+effv2MXDgQO1lxNCyNAUpJZ999hn79u3DzMyMqKgobty4Qb169ahatSrHjh3jxo0bODs7U7VqVSPZX/jftahTpw5169aldevWADRo0IBLly7xzjvv0KNHD62nKCPR0dFGoVq3bNmCp6cnFhYW+Pj4MGPGDL755hvMzc2zla/Nr6Stp6cnoaGh+TomLyZMmMC7776LTqfDwcEBZ2dnzM3N2bJlC9WrV8fV1TWLOuD8+fPZtm0brVq1Ys6cOXzwwQcsXboU0Evanjlz5rHamB0lzsknJiVQrox+pV0T891FbE3xIy0tjVmzZmlvom3atKFfv35YW1sXsWVPH6a2uB83Smq28KRms8u/atWqRmnmz59PjRo1CAsLIy0tzaQlqS+++CJ79uzhww8/pEKFCkgp+c9//pMlGt22bdvyzCsn/Pz8uHXrFkePHqVs2bLUq1dPU5McOXIky5cv559//tF6BnK6FhEREUbX0tramrCwMIKDg1m8eDGBgYFZJpVllLQFfVf9H3/8Qb169QC4c+cOu3fvxsvLS5O1NfQ4Zux9NPUezU9LvlatWkRGRmq/IyMjNSnajFhZWbFs2TKtburXr0+DBg1Ys2YNmzZtYtu2bSQmJnLv3j1effVV5s+fT1hYGK1a6WPA+fr60rVrVy2/JyVpW+LG5B+mpmnfLcS9IrSkeGJmZka7du2oX78+n332GaNGjVIO/ilFSc0+XqnZnPLPKPcK+lZ3zZo1MTMzY9WqVZrcqZeXF8uWLdPyyNhd//rrr9O9e3f69+9PSkoKXbp04fvvv9fq79y5c9y/f5927dqxZs0aUlNTiY6ONlKDzIvY2FiqV69O2bJl2bNnD1eu/E8LzNvbm+3bt3P48GHtxcLUa3H79m3S0tLw8fFh5syZ2vBgRjJK2t67d4/9+/dz9epVIiIiiIiI4LvvvtPU7jw8PDSBmNTUVFavXq0p/pk49XaqAAAgAElEQVR6jxpa8pk/mR08QM2aNbWudCklK1eupE+fPlnS/fvvv9pL69KlS2nXrh1WVlZ88cUXREZGEhERQUBAAB06dGD16tVYW1sTGxur2bpz506j3ownJWlb4lryAJduqChsBmJjY1m/fj0vvPCC9pbYsWNHOnbsqELRKoykZocMGUJQUBDvvPMOY8aMITU1lSFDhmQrNZuQkIAQwmg5kIHc0o0aNYo+ffrg5ORE165dCyQ1e/r0adq0aQPoJ+6tXr2a6tWrZ2uH4R7v378/NWvWZOzYsVpvQn6lZps1a5ZrmqNHj2abf3JyMubm5jg5OTF8+HDefvttfHx8WLlypVEddO3aldDQUFq0aEG5cuXo3r27kSb9Bx98QGxsLEOGDMHPz4+IiAhcXFyQUvLcc8+xceNGvL292b17N/b29tSpU0erp+x48803ee+99wCoXbs2mzdvplevXjg4ONCiRQsjydpy5crh6enJs88+q6nR5XQtMqvVRUVFMWLECK0H54svvshiS7du3RgyZAgAGzZsoEOHDkYS1X369OHjjz8mKSmJyZMnM3r0aJycnJBS0rVrV159VS9rYuo9ml8WLVrE8OHDefDgAd26daNbt24A2gvEW2+9xenTpxk2bBhCCJo1a8ZPP/2Ua55lypThxx9/xMfHBzMzM6ytrY16OA4cOMC0adMe2fa8KFSp2cKgZp06ctKgjwmLaI+D+Ta6lZ9Lo59vFbVZT5zk5GR+++03Nm/eTGJiIpaWlsydO5dy5coVtWlPNUpqVlESSUtLw8XFhbVr12Jra1soZXh7e/PVV18VWv4lidwkbUuM1GxhYW4uME+PW1/PvGABJEoyUkrCwsIICAjQus50Oh2+vr7KwSsUinxz6tQpevbsibe3d6E64NmzZxMdHa2cPE9W0rbEOfmMCEpWL8SjkpCQwOLFi7UZvTVr1mTgwIFKREahUBQYe3t7Ll26VOjl2NnZYWdnV+jllASe5IqXEu3knzYqVKhAQkICFhYW9OnThw4dOhgFl1AoFAqFIiMl3kNUdOxU1CYUGmlpaezbtw97e3uqV6+OmZkZI0eOpGLFilhZWRW1eQqFQqEo5pQ4J5+S+i+37/1vDeML7/kXoTWFx9mzZ/H39+fq1au4uLhoM6Cff/75IrZMoVAoFCWFEufky5olEnlbP3GjBIbez5M7d+4QGBjI4cOHAX2M+cIOe6hQKBSK0kmJ85ICSZrUv5tUN7tYxNY8PpKSkggKCmLixIkcPnyYcuXK8fLLLzNz5kzl5BX54mmRmnVwcNBCjBrEVoAcY6BnFM8ZOXIkp06demQ79u7daxRgZfHixUZyqo+DiIiIbCfX5rT9r7/+olWrVuh0Opo2bcq0adNYtmyZJrlbrlw5re4mTJjA8uXLEUJoQZIANm7ciBAiR7na9957zyh6okGiN3NgmswCWMuXLzeS8V25cqUmbOPs7KyJFD0K27dvx87OjkaNGmniS5m5cuUKHTt2xNHREQ8PDy3i3Z49e7R60ul0VKhQQROr+fbbb2nUqBFCCCMp3C1btjBlypRHtrvQKKh8XVF9bGpXkqN8j8tRvsdl+PA62Ur1lUSuX78uR44cKUeMGCG///57efv27aI2SVEAlNRs3jxuqdkzZ87IOnXyfhaYIr+akpKSLzumTp0q58yZk69j8svly5dls2bNTN7euHFjGRoaKqXUn8/JkyeN9meWcl22bJl0cHCQr7/+uratf//+0snJKdv6un37tmzVqpXRtkWLFsm2bdvKdu3aGW3PeC8ayjLI+G7btk06OzvLqKgoKaWUiYmJcsmSJVkrIB+kpKTIBg0ayIsXL8qkpCTp6OiY5fyllLJfv35y+fLlUkopd+3aJV999dUsae7cuSOtra3l/fv3pZRShoSEyMuXL2epv7S0NKnT6bR0j8pTLzUr0/4nsFJeJBShJY9OdHS01oqqWbMm/fv3Z8KECbz11ltZ4mErFAXhaZCavXfvnlHoZkOZMhdZVg8PD01e1tLSkg8//BAnJycOHjzI6tWradmyJTqdjjfffFMLS7t9+3ZcXFxwcnKiY8eOREREsHjxYubPn49Op2P//v1G5xEaGkrr1q1xdHTE29ubu3fvamV/8skntGzZksaNG7N//36AXCVq88PNmze1CILm5ubY29vneYxBWS05OZn4+HguXLiATqfLNu369euNYrCDPhb9vHnziIqKMooDnxtffPEFc+fO1ZQCy5cvz6hRo0w6NicOHTpEo0aNaNCgAeXKlWPAgAFGvTwGMsrGenp6Zptm3bp1dOvWTRMUcnZ21mLtZ0QIgYeHh6ZCWNwocU7eQJ1njxW1CQXm3r17LF++nEmTJmlxvkG/drJx48ZFaJnisbKvQuF8TMQgNdu7d2/ANKnZzPuzw9R0mQkJCWHBggWcO3cOX19fAgMDtX2BgYH4+voayZuGhoZy9OjRLKI6Bjw9PWnevDnt27dn5syZWfZv2LBBk2VduXJljk7z/v37tGrVirCwMKpWrcqaNWs4cOCApiJnEHYZNWoU69evJywsjLVr11KvXj3eeust3n//fUJDQ3F3dzfKd+jQoXz55ZccP34cBwcHpk+fru1LSUnh0KFDfPPNN9r26tWrs3PnTkJCQlizZg3jxo3Ldx0DvP/++9jZ2eHt7c0PP/xgJAyTE0IIOnXqRHBwMEFBQdo9kx2ZZWOvXbtGdHQ0LVu2pH///qxZs8YkO029j/z8/Iy60A2ffv36ZUkbFRVF7dq1td95ycaC/j6Ji4vjzp07RmkCAgIYOHCgSedikI0tjpS4iXdSltj3ElJSUti9ezdBQUE8ePAAc3Nzk8U3FApTeVqkZvfs2UO1atW4ePEiHTt2xMPDw2gM2FRZVnNzc02pbteuXRw9elSLef/gwQOqV6/OX3/9pQk/AbkK2YBeU+Lff/+lffv2AAwbNkwT4QHo27cvAK6urkRERAA5S9TmlylTpjB48GB27NjBf//7X/z9/bNIoGbHgAEDWLhwIbGxscybN88orn5GoqOjjeR716xZQ//+/bU8XnvttVznaORXNnbw4MEMHjw4X8fkxdy5cxk7dizLly+nXbt21KpVyygmf3R0NCdOnMiiApgT1atX5/r164/VxsdFiXPyaWl6UYPEFMs8UhYvTpw4QUBAANHR0QA0b96cgQMHat1qilJIu7xbUIXB0yY127BhQ2rUqMGpU6cKNEm1QoUK2gNeSsmwYcOyiKxs3rw53/nmhkGcxdzcnJSUFKBgErU50bBhQ0aPHs2oUaN47rnnuHPnTp5DgC1btuTEiRNUrFgx1x7F7GRj//nnH/z8/AC4fv0658+fx9bWFgsLCx4+fKiF3M5ONjanly8Dfn5+zJkzJ8v2Ro0aZZkYWKtWLa5du6b9zkk29oUXXtBa8vHx8axfv55nn31W2x8YGIi3tzdly5bNcmx2PCnZ2IJQ4prFAv0DpHKFf4rYEtM5ePAg8+fPJzo6mho1avDuu+/y/vvvKwevKFSeFqnZmzdvcvnyZerWrWu0vSCyrB07dmTdunVamTExMVy5coXWrVuzb98+Ll++rG0HssjMGqhcuTLW1tZaF+6qVau0Vn1O5CRRm1+2bt2qzfU5f/485ubmRg4sN2bPnp1jC95ARtnYc+fOER8fT1RUlCYb++mnn2qyse3bt2f16tWA/n4LDAzUZGM//fRTxo8fzz//6J/lDx8+ZOnSpVnKGzx4cLaysdnN/Hdzc+P8+fNcvnyZhw8fEhAQkO3Qg0EeF/RzA1577TWj/f7+/iZ31RvqobiGFy9xTr6cmf7mrVHhSh4pixbDPxmAi4sLNjY29O/fnxkzZuDk5JTvLiuFoiBklJq1sLAgKCiImTNnYmdnh4ODA25ubtlKzTZt2pTmzZtnG9M8t3SjRo3i999/1yaxFURqdtCgQbRp0wYHBwf69euXrRMF/Zi8TqfD09OT2bNnU6NGDaP9BsEVe3t7hg4dmqssqwF7e3tmzpxJ586dcXR0xMvLS+ueXrJkCX379sXJyUnreejVqxcbNmzQJt5lZMWKFYwfPx5HR0dCQ0PzXGb19ttvs2LFCpycnDhz5kyudWfg7Nmz2NjYaJ+1a9eyatUq7Ozs0Ol0mmxtZnnYnOjWrZvmhHOiR48eWve/v78/3t7eRvt9fHw0J79gwQJ++eUXdDodrVu35pVXXtGGXrp3787YsWPp1KkTzZo1w8XFhXv37plkZ06UKVOGb7/9li5dutC0aVP69++vSQhPmTKFTZs2AfqJoHZ2djRu3JgbN24wceJELY+IiAiuXbuW5aVs4cKF2NjYEBkZiaOjIyNHjtT27dmzhx49ejyS7YVFiZOarVOrhuzq/hvt6qyi9e1lxU5mNi0tjQMHDrBz504mTJigzcxMS0tT+u5PAUpqVvE00LZtW7Zs2WJyD0Fp5saNGwwaNIhdu3Y9lvyeeqlZAyn3i19L+MKFC/j5+XHlir6X4cCBA9qkJ+XgFQpFaWHevHlcvXpVOXng6tWrzJs3r6jNyJES6+TLULDxqsIgJiaGdevW8ddffwFgbW3NK6+8QqtWrYrYMoVCoXj8qGfb/zCsxCiulFgnD8VDgW7//v34+fnx8OFDypQpQ9euXenRo4c2e1ahUCgUiqKiRDv54qBAV6VKFR4+fIirqyv9+/c3Wj+qUCgUCkVRUuKcfHmzop0oeO3aNU6fPq2FBm3WrBnTp083irKkUCgUCkVxoMQ5+aQUfTe4THmy5cbHx7Nhwwb27t2LlBJbW1st+pVy8AqFQqEojpQ4J5+apjc5IemZJ1Neaip79+5l48aN3L9/HzMzMzp06KC65RUKhUJR7Clx67rMzfSz6muVCS/0sk6dOsW0adPw8/Pj/v37mjbz4MGDs+gkKxTFhadBTz4jycnJTJgwAVtbW1xcXGjTpg2//voroI/AZ4hLD3plseHDhwN6bXMzMzOOHz+u7W/evLkWS76g7N27l549e2bZvmXLFpydnXFycsLe3p4ffviBWbNmaYIrhuum0+lYuHAh06ZNQwihRZcD+OabbxBCaAp6menXr59RAKPQ0FCEEGzfvl3blp0OfUb1PNDHdm/SpAk6nQ43NzdWrlxZ4PowsGLFCmxtbbG1tWXFihXZpgkLC9MCIfXq1csoOM7x48dp06YNzZo1w8HBIYvwTu/evY3O66OPPmL37t2PbHdJp8Q5eQNlZVKhl3Hw4EGioqKoVq0aY8eO5aOPPsLGxqbQy1UoHgVD7Prw8HCqVKmiSbo+ePCA3r17M2HCBM6ePUtYWBh//vknixYtAvSqYGPHjmX16tWcOnWKI0eO0KhRoyz5m5ouJwyx2h8XkydPJjo6mvDwcEJCQti4caNRlLyjR49y6tSpbI+1sbFh1qxZ+SovO7nRvEhOTuaNN95g8+bNhIWFcezYMTw8PJg4caIWptVw3UJDQzUFOgcHBwICArR81q5dq0Vwy8zJkydJTU2lQYMG2jZ/f3/atm2rRaAzhcWLF7Nz505NCXDXrl08atC0mJgYpk+fzt9//82hQ4eYPn26Jr2bkZEjRzJ79mxOnDiBt7e3FrM+JSWFV199lcWLF3Py5En27t1rFFf+l19+ydLweueddzQZ5aeZktddn6q/sGnJjz8YTmJiInfv3tVCbPr4+PD888/TuXNnk4UKFAoDbww4USj5LglwMDltmzZttJZqTnryHh4ejBkz5rHpyffs2VOTAbW0tCQ+Pp69e/cyefJkrK2tOXPmDH379qV27dqMGTMG0LckLS0t+eijj5gzZw6BgYEkJSXh7e1tJNGamYSEBH788UcuX76sLVutUaOGpooG8OGHHzJr1ixNQCUjPXv2ZN++fZw9exY7OzuT6zW/xMXFkZKSoonElC9f3qTyXn75ZYKCgpg0aRIXL16kcuXKOT6L/Pz86NOnj/ZbSsnatWvZuXMn7u7uJCYmmiR68/nnn7N3716srKwAsLKyYtiwYaacZo4EBwfj5eWlqfd5eXmxffv2LPHhz507p4W99fLyokuXLsyYMYMdO3bg6OioiSJlFNuJj4/n66+/ZsmSJUbXvW7duty5c4d//vmH559//pHsL8mUuJb8wxTDTfr4nHxaWhp//vknn376KYsWLdKEIZ599ll69OihHLyiRFLa9eRBH2WyTp06mkPKjv79+xMSEmLU7W3AzMyMjz/+OE9RlkelSpUq9O7dm7p16zJw4ED8/Pw0gZTcsLKyonbt2oSHhxMQEKDFzM+OzDrvf/75J/Xr16dhw4Z4eHiwdevWPMu7d+8ecXFxRr0BOTFnzpxsdd4NvRAZMVXnvVmzZgQFBQH6XguDoty5c+cQQtClSxdcXFz46quvtGMmT57Mhx9+qIUQz4iLiwsHDhzI81xKMyWuJW9upu/qe84sq3BGQbh06RL+/v5cvHgR0P8z3rt3D2tr68eSv+LpJT8t7sfJ06Inbyrm5uaMHz+eL774gm7dumXZP2jQIGbNmqUpzGXHmDFjNGdx/fp1dDodoJfLzShukhtLly7lxIkT/Pbbb8ydO5edO3eyfPnyPI8bMGAAAQEBBAcHs2vXLpYtW5Ztusw67/7+/gwYMEDLY+XKlfj4+OQojpVf0azx48czfvz4fB2TFz///DPjxo1jxowZ9O7dW5OoTUlJ4Y8//uDw4cNUrFiRjh074urqStWqVbl48SLz58/Pdi5FcdZ5f1KUOCdvwJyHeSfKhX///Zf169dr/7iVK1emX79+tGnTRsWZV5RoniY9+UaNGnH16lXu3buXa2t+yJAhfPHFF9nKgZYpU4YPP/yQL7/8MsfjDfMaQD8mHxoaapJ9mXFwcMDBwYEhQ4ZQv359k5x8z549GT9+PC1atMj1HDPqvKemprJ+/XqCgoKYNWsWUkru3LlDXFwcVatWzTIeHhMTQ/369bGyssLS0pJLly7l2ZqfM2dOtkMg7dq1Y+HChUbbatWqpSnXgV7n3cPDI8uxTZo0YceOHYC+9W7ofbCxsaFdu3aaFn337t0JCQnB0tKSI0eOUK9ePVJSUrh58yYeHh5aWcVZ5/2JIaUsUZ8aVRvLUb7H5ekvR8qCkpqaKj/55BM5YsQIOWrUKBkYGCgTEhIKnJ9CYeDUqVNFbYKsVKmS9j0kJETWqVNHJicny4SEBFm/fn25c+dOKaWUCQkJskePHnLhwoVSSinDwsJkw4YN5dmzZ6WU+v+T77//Pkv+uaWbMWOG/Pjjj6WUUm7YsEHqHzFS7tmzR/bo0cMon/DwcNmmTRtpa2srr1+/LqWUMjg4WLZs2VLGxcVJKaWMjIyUN27cyPV8x48fL4cPHy6TkpKklFLevHlTBgYGSimlrFu3rrx165aUUsrvvvtO1q5dWw4bNkxKKeWyZcvkmDFjpJRSJiUlyYYNG8rq1avLy5cv51pe3bp1c92f3bnGxcXJPXv2aL937twpmzVrZpQm43WTUsqpU6fKOXPmSCml9Pf3l0ePHpVSStm+fXt5+PDhLOX6+vpq1zY4OFh27tzZaP/QoUPlihUrpJRSurq6yl27dkkppbxz5460tbWVFy5ckFLq66lr164yNjZWs91wXEG5c+eOrFevnoyJiZExMTGyXr168s6dO1nSGa51amqqHDJkiPzpp5+klFLGxMRIZ2dnef/+fZmcnCw7duwot2zZYnTs5cuXs9Rpz5495cGDBx/J9idNds8Q4IgsoM8ssU3W50cuyld6KaXWwjAzM6Nbt27odDpmzJjBK6+8ot72FKWS0qwnb2DmzJk899xz2Nvb07x5c3r27Jlti/f111/PcWZ/uXLlGDduHDdv3sy1LFPZtWuXkc77sWPH+OqrrzSd96lTp5rUijcwYMAAXFxcck2TH533lStXMmPGDHQ6HR06dGDq1Kk0bNgQgNGjR+Pp6YmbmxvNmzfH3d39kXs3q1SpwuTJk3Fzc8PNzY0pU6Zok/BGjhypLQn09/encePGNGnShBdeeIERI0YAetGvDz74ADc3N3Q6HS4uLnnqtycnJ3PhwgVatCiQQmupocTpyT9fzU727rSOrxY14dkqpk2Ii4qKIiAgAFtbW20SkpQy32NQCkVeKD15RVHx4MEDPD09OXDgAObm5kVtTpGzYcMGQkJCmDFjRlGbki+eej15Q8Q7U7h//z5BQUHs3r2btLQ0rl+/Trdu3Shbtqxy8AqFolRhYWHB9OnTiYqKok6dOkVtTpGTkpJSpEGaigslzskbqFAx5+6jtLQ0fv/9dzZs2EB8fDxCCDw8PHj55ZfVcjiFQlFq6dKlS1GbUGx45ZVXitqEYkHJdfIVsu+OiouLY86cOURGRgJgZ2fHwIED1Zut4omhhoIUCkVBKIzh8xLp5CU5h8W0tLSkUqVKVK1aFV9fX1xdXdUDV/HEqFChAnfu3KFq1arqvlMoFCYj05c5mhKVMD+USCePxf8EN5KSkvj1119p1aoVNWvWRAjBG2+8QaVKlbRACgrFk8LGxobIyEhu3bpV1KYoFIoSRoUKFR67PkqhOnkhRFdgAWAOLJVSzs60vzywEnAF7gC+UsqIvPJ9/rnqSCk5dOgQgYGB3L17l8uXL/P+++8DqGh1iiKjbNmyWlQ3hUKhKGoKzckLIcyB7wAvIBI4LITYJKXMKAf1OnBXStlICDEA+BLIOTizIe9yN/niiyAtFnXdunXzXDOpUCgUCsXTRmG25FsCF6SUlwCEEAFAHyCjk+8DTEv/vg74VgghZC6zD6S4R+T1VYBevKFv3760bdtWhaJVKBQKhSIThenkawHXMvyOBFrllEZKmSKEiAWqArdzzFUkIoQZnTt70atXr2yVhxQKhUKhUJSQiXdCiDeAN9J/Jv3880/hP//8U1GaVNqpRm4vWorHharnwkfVceGj6rjwsSvogYXp5KOA2hl+26Rvyy5NpBCiDFAZ/QQ8I6SUS4AlAEKIIwUN76cwDVXHTwZVz4WPquPCR9Vx4SOEOFLQYwtzIPswYCuEqC+EKAcMADZlSrMJGJb+vR+wO7fxeIVCoVAoFKZTaC359DH2sUAw+iV0P0spTwoh/g+9bN4m4CdglRDiAhCD/kVAoVAoFArFY6BQx+SllNuAbZm2TcnwPRHIb4DhJY/BNEXuqDp+Mqh6LnxUHRc+qo4LnwLXcYmTmlUoFAqFQmEaanG5QqFQKBSllGLr5IUQXYUQZ4UQF4QQE7LZX14IsSZ9/99CiHpP3sqSjQl1/IEQ4pQQ4rgQYpcQom5R2FmSyauOM6TzEUJIIYSapVwATKlnIUT/9Pv5pBDiv0/axpKOCc+LOkKIPUKIY+nPjO5FYWdJRgjxsxDiphAiPIf9QgixMP0aHBdCuOSZqZSy2H3QT9S7CDQAygFhgH2mNG8Di9O/DwDWFLXdJeljYh17AhXTv49Wdfz46zg93TPAPuAvoEVR213SPibey7bAMcA6/Xf1ora7JH1MrOMlwOj07/ZARFHbXdI+QDvABQjPYX934FdAAK2Bv/PKs7i25LWQuFLKh4AhJG5G+gAr0r+vAzoKpe2ZH/KsYynlHillQvrPv9DHOlCYjin3McAM9LoNiU/SuFKEKfU8CvhOSnkXQEp58wnbWNIxpY4lYJX+vTJw/QnaVyqQUu5Dv9IsJ/oAK6Wev4BnhRA1c8uzuDr57ELi1sopjZQyBTCExFWYhil1nJHX0b9BKkwnzzpO726rLaXc+iQNK2WYci83BhoLIQ4IIf5KV8hUmI4pdTwNeFUIEYl+VdU7T8a0p4r8PrdLRlhbRdEihHgVaAG0L2pbShNCCDPga2B4EZvyNFAGfZe9B/oeqX1CCAcp5b9FalXpYiCwXEo5TwjRBn0MlOZSyrSiNuxppri25PMTEpfcQuIqcsSUOkYI0QmYCPSWUiY9IdtKC3nV8TNAc2CvECIC/RjbJjX5Lt+Yci9HApuklMlSysvAOfROX2EaptTx60AggJTyIFABfVx7xePDpOd2Roqrk1chcQufPOtYCOEM/IDewasxzPyTax1LKWOllNWklPWklPXQz3voLaUscJzqpxRTnhcb0bfiEUJUQ999f+lJGlnCMaWOrwIdAYQQTdE7+VtP1MrSzyZgaPos+9ZArJQyOrcDimV3vVQhcQsdE+t4DmAJrE2f03hVStm7yIwuYZhYx4pHxMR6DgY6CyFOAanAeCml6vkzERPr+EPgRyHE++gn4Q1XDa/8IYTwR/8yWi19bsNUoCyAlHIx+rkO3YELQAIwIs881TVQKBQKhaJ0Uly76xUKhUKhUDwiyskrFAqFQlFKUU5eoVAoFIpSinLyCoVCoVCUUpSTVygUCoWilKKcvEJRBAghUoUQoRk+9XJJG/8YylsuhLicXlZIekSy/OaxVAhhn/79s0z7/nxUG9PzMdRLuBBisxDi2TzS65TamUKRM2oJnUJRBAgh4qWUlo87bS55LAe2SCnXCSE6A3OllI6PkN8j25RXvkKIFcA5KeWsXNIPR6/cN/Zx26JQlAZUS16hKAYIISyFELvSW9knhBBZ1OqEEDWFEPsytHTd07d3FkIcTD92rRAiL+e7D2iUfuwH6XmFCyHeS99WSQixVQgRlr7dN337XiFECyHEbMAi3Q6/9H3x6X8DhBA9Mti8XAjRTwhhLoSYI4Q4nK6D/aYJ1XKQdPENIUTL9HM8JoT4Uwhhlx557f8A33RbfNNt/1kIcSg9bXaqfwrFU0OxjHinUDwFWAghQtO/XwZeAbyllPfSw67+JYTYlCli2CAgWEo5SwhhDlRMTzsJ6CSlvC+E+AT4AL3zy4lewAkhhCv6iFmt0OtT/y2E+B29Zvh1KWUPACFE5YwHSyknCCHGSil12eS9BugPbE13wh2B0ejjmsdKKd2EEOWBA0KIHelx5LOQfn4d0Ue2BDgDuKdHXusEfC6l9BFCTCFDS14I8Tn6ENevpXf1HxJC/BPv7xgAAAKLSURBVCalvJ9LfSgUpRbl5BWKouFBRicphCgLfC6EaAekoW/B1gD+yXDMYeDn9LQbpZShQoj2gD16pwlQDn0LODvmCCEmoY8n/jp6J7rB4ACFEL8A7sB2YJ4Q4kv0Xfz783FevwIL0h15V2CflPJB+hCBoxCiX3q6yugFYjI7ecPLTy3gNLAzQ/oVQghb9CFTy+ZQfmegtxDio/TfFYA66XkpFE8dyskrFMWDwcBzgKuUMlnoVekqZEwgpdyX/hLQA1guhPgauAvslFIONKGM8VLKdYYfQoiO2SWSUp4Tep377sBMIcQuKWVuPQMZj00UQuwFugC+QIChOOAdKWVwHlk8kFLqhBAV0cdJHwMsBGYAe6SU3umTFPfmcLwAfKSUZ02xV6Eo7agxeYWieFAZuJnu4D2BupkTCCHqAjeklD8CSwEX9Mp1LwkhDGPslYQQjU0scz/wshCiohCiEuAN7BdCvAAkSClXoxcpcsnm2OT0HoXsWIN+GMDQKwB6hz3acIwQonF6mdkipUwAxgEfiv9JSRskNYdnSBqHXrLXQDDwjkjv1hB6JUWF4qlFOXmFonjgB7QQQpwAhqIfg86MBxAmhDiGvpW8QEp5C73T8xdCHEffVd/ElAKllCHAcuAQ8DewVEp5DHBAP5Ydil4Fa2Y2hy8Bjhsm3mViB9Ae+E1K+TB921LgFBAihAhHL2Gca09iui3HgYHAV8AX6eee8bg9gL1h4h36Fn/ZdNtOpv9WKJ5a1BI6hUKhUChKKaolr1AoFApFKUU5eYVCoVAoSinKySsUCoVCUUpRTl6hUCgUilKKcvIKhUKhUJRSlJNXKBQKhaKUopy8QqFQKBSlFOXkFQqFQqEopfw/VN4B0vkl9IkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "linewidth = 2\n",
    "plt.figure(figsize=(8, 5))\n",
    "\n",
    "plt.plot(fprB, tprB, color='#db6114', lw=linewidth, label='ROC Curve LSTM Stochastic GD (AUC = %0.3f)' % auc(fprB, tprB))\n",
    "plt.plot(fprMB, tprMB, color='#1565c0', lw=linewidth, label='ROC Curve LSTM Mini Batch(AUC = %0.3f)' % auc(fprMB, tprMB))\n",
    "plt.plot(fpr3L, tpr3L, color='#2e7d32', lw=linewidth, label='ROC Curve LSTM Stacked Layers (AUC = %0.3f)' % auc(fpr3L, tpr3L))\n",
    "plt.plot(fprBi, tprBi, color='#ffc107', lw=linewidth, label='ROC Curve Bidirectional LSTM (AUC = %0.3f)' % auc(fprBi, tprBi))\n",
    "plt.plot(fprCNN, tprCNN, color='#6557d2', lw=linewidth, label='ROC Curve  CNN + LSTM (AUC = %0.3f)' % auc(fprCNN, tprCNN))\n",
    "plt.plot([0, 1], [0, 1], color='#616161', lw=linewidth, linestyle='--')\n",
    "\n",
    "plt.xlim([0.0, 1.0])\n",
    "plt.ylim([0.0, 1.05])\n",
    "\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.title('ROC Curve Plots')\n",
    "plt.legend(loc=\"lower right\")\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "HihEst91ZtXJ"
   },
   "source": [
    "As we can see, all models show approximately the same result, but so far the best is the model with Stacked Layers (90.4%), so we gonna use it. Here we are finished, then we move on to the development of a web application."
   ]
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "Part_III_Training_and_Development_of_Different_LSTM_Models.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}