experiment-files/final_classification.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Extract data from jsons and insert into CSV"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import csv\n",
    "import json\n",
    "\n",
    "\n",
    "folder_path = '../receipts/json'\n",
    "\n",
    "items = []\n",
    "\n",
    "for file in os.listdir(folder_path):\n",
    "    receipt_path = os.path.join(folder_path, file)\n",
    "\n",
    "    with open(receipt_path) as receipt_data:\n",
    "        receipt = json.load(receipt_data)\n",
    "\n",
    "    item_list = receipt['itemList'] + receipt['discountList']\n",
    "\n",
    "    for item in item_list:\n",
    "        items.append([receipt['storeName'],item['itemName'],item['category']])\n",
    "\n",
    "with open('../receipts/categories.csv', 'w', newline='') as file:\n",
    "    header = [['store_name', 'item_name', 'category']]\n",
    "    writer = csv.writer(file)\n",
    "    writer.writerows(header+items)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Clean and prep data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import string\n",
    "\n",
    "def lower(text):\n",
    "    result = \" \".join([word.lower() for word in text.split()])\n",
    "    return result\n",
    "\n",
    "def remove_punctuation(product):\n",
    "    text = str.maketrans((string.punctuation + '£'), ' '*len((string.punctuation + '£')))\n",
    "    return product.translate(text)\n",
    "\n",
    "def remove_nums(product):\n",
    "    text = str.maketrans(string.digits, ' '*len(string.digits))\n",
    "    return product.translate(text)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(657, 3)\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>store_name</th>\n",
       "      <th>item_name</th>\n",
       "      <th>category</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>TESCO</td>\n",
       "      <td>SEASONAL VEG</td>\n",
       "      <td>groceries</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>LiDL</td>\n",
       "      <td>Frijj Cookie Dough</td>\n",
       "      <td>drinks</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Sainsbury's</td>\n",
       "      <td>JS DBLE CHC COOKIES</td>\n",
       "      <td>snacks</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>TESCO</td>\n",
       "      <td>T.ZIPSEAL BAGS</td>\n",
       "      <td>homeware</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>TESCO</td>\n",
       "      <td>ICE CREAM</td>\n",
       "      <td>snacks</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>LiDL</td>\n",
       "      <td>Mackeral Sunflower</td>\n",
       "      <td>groceries</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>Sainsbury's</td>\n",
       "      <td>BLT SW</td>\n",
       "      <td>groceries</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>ALDI</td>\n",
       "      <td>THYME CUT</td>\n",
       "      <td>groceries</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>lidl</td>\n",
       "      <td>Green Tee Honey</td>\n",
       "      <td>drinks</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>TESCO</td>\n",
       "      <td>CC Kitchen towels</td>\n",
       "      <td>cleaning &amp; laundry</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    store_name            item_name            category\n",
       "0        TESCO         SEASONAL VEG           groceries\n",
       "1         LiDL   Frijj Cookie Dough              drinks\n",
       "2  Sainsbury's  JS DBLE CHC COOKIES              snacks\n",
       "3        TESCO       T.ZIPSEAL BAGS            homeware\n",
       "4        TESCO            ICE CREAM              snacks\n",
       "5         LiDL   Mackeral Sunflower           groceries\n",
       "6  Sainsbury's               BLT SW           groceries\n",
       "7         ALDI            THYME CUT           groceries\n",
       "8         lidl      Green Tee Honey              drinks\n",
       "9        TESCO    CC Kitchen towels  cleaning & laundry"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "path = '../receipts/categories.csv'\n",
    "dataset = pd.read_csv(path, header=0, index_col=0)\n",
    "\n",
    "dataset = dataset.sample(657).reset_index()\n",
    "\n",
    "print(dataset.shape)\n",
    "dataset.head(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## NLTK preprocesing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "from nltk.corpus import stopwords\n",
    "from nltk.stem import PorterStemmer\n",
    "\n",
    "stop = stopwords.words('english')\n",
    "stems = PorterStemmer()\n",
    "\n",
    "def remove_stopwords_nltk(text):\n",
    "    text = [word for word in text.split() if (word.lower() not in stop and len(word) > 1)]\n",
    "    return \" \".join(text)\n",
    "\n",
    "def stemmer_nltk(text):\n",
    "    stemmed = [stems.stem(word) for word in text.split()]\n",
    "    return \" \".join(stemmed)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "    store_name            item_name            category\n",
      "0        tesco         seasonal veg           groceries\n",
      "1         lidl   frijj cookie dough              drinks\n",
      "2  sainsbury s  js dble chc cookies              snacks\n",
      "3        tesco         zipseal bags            homeware\n",
      "4        tesco            ice cream              snacks\n",
      "5         lidl   mackeral sunflower           groceries\n",
      "6  sainsbury s               blt sw           groceries\n",
      "7         aldi            thyme cut           groceries\n",
      "8         lidl      green tee honey              drinks\n",
      "9        tesco       kitchen towels  cleaning & laundry\n"
     ]
    }
   ],
   "source": [
    "stop = ['cc', 'wr', 'tesco', 'aa', 'ess']\n",
    "\n",
    "units_solid = ['kg', 'g', 'gram', 'grams']\n",
    "units_liquid = ['ml', 'l', 'cl', 'ltr']\n",
    "units_length = ['m', 'mm', 'cm', 'km', 'metre']\n",
    "units = ['pk', 'sm', 'lg', 'large', 'extra', 'sng']\n",
    "\n",
    "def remove_stopwords(text):\n",
    "    text = [word for word in text.split() if (word.lower() not in stop and len(word) > 1)]\n",
    "    return \" \".join(text)\n",
    "\n",
    "def replace_units(text):\n",
    "    words = []\n",
    "    for word in text.split():\n",
    "        if word in units_solid:\n",
    "            words.append('<UNIT_SOLID>')\n",
    "        elif word in units_liquid:\n",
    "            words.append('<UNIT_LIQUID>')\n",
    "        elif word in units:\n",
    "            words.append('<UNIT>')\n",
    "        else:\n",
    "            words.append(word)\n",
    "    return \" \".join(words)\n",
    "\n",
    "dataset['item_name'] = dataset['item_name'].astype(str)\n",
    "dataset['item_name'] = dataset['item_name'].apply(lower)\n",
    "dataset['item_name'] = dataset['item_name'].apply(remove_punctuation)\n",
    "dataset['item_name'] = dataset['item_name'].apply(remove_nums)\n",
    "dataset['item_name'] = dataset['item_name'].apply(replace_units)\n",
    "dataset['item_name'] = dataset['item_name'].apply(remove_stopwords)\n",
    "#dataset['item_name'] = dataset['item_name'].apply(stemmer_nltk)\n",
    "\n",
    "dataset['store_name'] = dataset['store_name'].astype(str)\n",
    "dataset['store_name'] = dataset['store_name'].apply(lower)\n",
    "dataset['store_name'] = dataset['store_name'].apply(remove_punctuation)\n",
    "\n",
    "print(dataset.head(10))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = dataset.to_csv('../receipts/products_list_final.csv', encoding='utf-8')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.svm import LinearSVC\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "from sklearn.feature_extraction.text import TfidfTransformer\n",
    "import numpy as np\n",
    "from keras.utils import np_utils\n",
    "\n",
    "\n",
    "def split_dataset(df):\n",
    "    train, test, ytrain, ytest = train_test_split(df['item_name'].astype(str), df['category'], test_size=0.2, random_state=88)\n",
    "\n",
    "    return train, test, ytrain, ytest\n",
    "\n",
    "\n",
    "def create_labels(train_labels, test_labels, labels):\n",
    "    encoder = LabelEncoder()\n",
    "\n",
    "    encoder.fit(labels)\n",
    "\n",
    "    y_train = encoder.transform(train_labels)\n",
    "    y_val = encoder.transform(test_labels)\n",
    "\n",
    "    return y_train, y_val, encoder\n",
    "    \n",
    "\n",
    "def create_one_hot_labels(Y_train, Y_test, labels):\n",
    "    encoder = LabelEncoder()\n",
    "    encoder.fit(labels)\n",
    "\n",
    "    Y_train = encoder.transform(Y_train)\n",
    "    Y_test = encoder.transform(Y_test)\n",
    "\n",
    "    n_classes = np.max(Y_train) + 1\n",
    "\n",
    "    Y_train = np_utils.to_categorical(Y_train, n_classes)\n",
    "    Y_test = np_utils.to_categorical(Y_test, n_classes)\n",
    "\n",
    "    return Y_train, Y_test, n_classes, encoder\n",
    "\n",
    "\n",
    "def transform_data(train, val):\n",
    "    vectorizer = CountVectorizer(max_features=5000)\n",
    "    transformer = TfidfTransformer()\n",
    "\n",
    "    X_train = vectorizer.fit_transform(train) # BoW\n",
    "    X_train = transformer.fit_transform(X_train) # TF-IDF\n",
    "\n",
    "    X_val = vectorizer.transform(val)\n",
    "    X_val = transformer.transform(X_val)\n",
    "\n",
    "    print(X_train.shape)\n",
    "\n",
    "    return X_train, X_val\n",
    "\n",
    "\n",
    "def get_svm_model(x,y):\n",
    "    svm = LinearSVC(max_iter=1000).fit(x,y)\n",
    "    return svm\n",
    "\n",
    "\n",
    "def get_acc(m, x, y):\n",
    "    predictions = m.predict(x)\n",
    "    acc = np.mean(predictions == y)*100\n",
    "    return acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "custom_data = pd.read_csv(\"../receipts/products_list_final.csv\", index_col=0, header=0, names=[\"store_name\", \"item_name\", \"category\"])\n",
    "custom_labels = custom_data.category.unique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>store_name</th>\n",
       "      <th>item_name</th>\n",
       "      <th>category</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>tesco</td>\n",
       "      <td>seasonal veg</td>\n",
       "      <td>groceries</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>lidl</td>\n",
       "      <td>frijj cookie dough</td>\n",
       "      <td>drinks</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>sainsbury s</td>\n",
       "      <td>js dble chc cookies</td>\n",
       "      <td>snacks</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>tesco</td>\n",
       "      <td>zipseal bags</td>\n",
       "      <td>homeware</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>tesco</td>\n",
       "      <td>ice cream</td>\n",
       "      <td>snacks</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>652</th>\n",
       "      <td>tesco</td>\n",
       "      <td>yoghuts</td>\n",
       "      <td>snacks</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>653</th>\n",
       "      <td>lidl</td>\n",
       "      <td>organic vine tomato</td>\n",
       "      <td>groceries</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>654</th>\n",
       "      <td>lidl</td>\n",
       "      <td>bells ibuprofen</td>\n",
       "      <td>health &amp; pharmacy</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>655</th>\n",
       "      <td>sushi jun</td>\n",
       "      <td>unagi cucumber roll</td>\n",
       "      <td>eating out</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>656</th>\n",
       "      <td>tesco</td>\n",
       "      <td>ice cream</td>\n",
       "      <td>snacks</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>657 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      store_name            item_name           category\n",
       "0          tesco         seasonal veg          groceries\n",
       "1           lidl   frijj cookie dough             drinks\n",
       "2    sainsbury s  js dble chc cookies             snacks\n",
       "3          tesco         zipseal bags           homeware\n",
       "4          tesco            ice cream             snacks\n",
       "..           ...                  ...                ...\n",
       "652        tesco              yoghuts             snacks\n",
       "653         lidl  organic vine tomato          groceries\n",
       "654         lidl      bells ibuprofen  health & pharmacy\n",
       "655    sushi jun  unagi cucumber roll         eating out\n",
       "656        tesco            ice cream             snacks\n",
       "\n",
       "[657 rows x 3 columns]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "custom_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(525, 572)\n"
     ]
    }
   ],
   "source": [
    "X_train, X_test, Y_train, Y_test = split_dataset(custom_data)\n",
    "\n",
    "Y_train, Y_test, encoder = create_labels(Y_train, Y_test, custom_labels)\n",
    "\n",
    "X_train, X_test = transform_data(X_train, X_test)\n",
    "\n",
    "svm_model = get_svm_model(X_train, Y_train)\n",
    "\n",
    "svm_acc = get_acc(svm_model, X_test, Y_test)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "79.54545454545455"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "svm_acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(525, 563)\n",
      "[[  5   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   9   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   5   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   8   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0  67   0   0   0   0   2   0   0   1   0   0   1]\n",
      " [  0   0   0   0   0  21   0   0   0   2   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   3   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0  21   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   1   0   2   7   1   0   0 338   0   0   1   1   1   8]\n",
      " [  0   0   0   0   0   0   0   0   0   0   3   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0  10   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   3   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   5   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0  18   0]\n",
      " [  0   1   0   0   0   0   0   0   1   1   0   0   0   0   1 109]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1440x1440 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.metrics import confusion_matrix\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "X_train, X_test, Y_train, Y_test = split_dataset(custom_data)\n",
    "\n",
    "X_test = custom_data['item_name'].astype(str)\n",
    "Y_test = custom_data['category']\n",
    "\n",
    "Y_train, Y_test, encoder = create_labels(Y_train, Y_test, custom_labels)\n",
    "\n",
    "X_train, X_test = transform_data(X_train, X_test)\n",
    "\n",
    "svm_model = get_svm_model(X_train, Y_train)\n",
    "\n",
    "predictions = svm_model.predict(X_test)\n",
    "\n",
    "predictions = encoder.inverse_transform(predictions)\n",
    "Y_test = encoder.inverse_transform(Y_test)\n",
    "\n",
    "labels = encoder.classes_\n",
    "\n",
    "cm = confusion_matrix(predictions, Y_test)\n",
    "print(cm)\n",
    "fig, ax = plt.subplots( figsize= (20, 20))\n",
    "\n",
    "ax.imshow(cm)\n",
    "\n",
    "ax.set_xticks(range(0, 16))\n",
    "ax.set_yticks(range(0, 16))\n",
    "\n",
    "ax.set_yticklabels(labels)\n",
    "ax.set_xticklabels(labels, rotation='vertical')\n",
    "\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(525, 563)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array(['eating out', 'groceries', 'groceries', 'snacks', 'groceries',\n",
       "       'drinks', 'drinks', 'snacks'], dtype=object)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train, X_test, Y_train, Y_test = split_dataset(custom_data)\n",
    "\n",
    "Y_train, Y_test, encoder = create_labels(Y_train, Y_test, custom_labels)\n",
    "\n",
    "X_train, X_test = transform_data(X_train, ['fried chicken', 'poems', 'biscuits', 'ice cream', 'vodka', 'lemon tea', 'lemon iced tea', 'lemon ice tea'])\n",
    "\n",
    "svm_model = get_svm_model(X_train, Y_train)\n",
    "\n",
    "#svm_acc = get_acc(svm_model, X_test, Y_test)\n",
    "\n",
    "preds = svm_model.predict(X_test)\n",
    "\n",
    "encoder.inverse_transform(preds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "import imblearn\n",
    "from imblearn.over_sampling import SMOTE, RandomOverSampler\n",
    "from imblearn.pipeline import make_pipeline\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "oversample = SMOTE()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = custom_data['item_name'].astype(str)\n",
    "Y_train = custom_data['category']\n",
    "\n",
    "encoder = LabelEncoder()\n",
    "\n",
    "Y_train = encoder.fit_transform(Y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(657,)"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Y_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(657,)"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sm = SMOTE(random_state=42) \n",
    "rm = RandomOverSampler(random_state=42)\n",
    "my_pipe = make_pipeline(rm, sm) \n",
    "X_res, Y_res = my_pipe.fit_resample(X_train, Y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:`period` argument is deprecated. Please use `save_freq` to specify the frequency in number of batches seen.\n"
     ]
    }
   ],
   "source": [
    "from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn.model_selection import train_test_split\n",
    "import numpy as np\n",
    "from keras.utils import np_utils\n",
    "from keras.layers import TextVectorization\n",
    "import keras.layers as l\n",
    "import tensorflow as tf\n",
    "from gensim.models.fasttext import FastText\n",
    "from gensim.test.utils import datapath\n",
    "\n",
    "from keras.models import Sequential\n",
    "\n",
    "from keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau\n",
    "\n",
    "mc = ModelCheckpoint(\n",
    "    'final.h5',\n",
    "    monitor='val_accuracy',\n",
    "    verbose=0,\n",
    "    save_best_only=True,\n",
    "    mode='auto',\n",
    "    save_weights_only=False,\n",
    "    period=1\n",
    ")\n",
    "\n",
    "es = EarlyStopping(\n",
    "    monitor='val_accuracy',\n",
    "    min_delta=0.001,\n",
    "    patience=100,\n",
    "    mode='auto'\n",
    ")\n",
    "\n",
    "rlr = ReduceLROnPlateau(\n",
    "    monitor='val_accuracy',\n",
    "    factor=0.8,\n",
    "    patience=50,\n",
    "    min_delta=0.001,\n",
    "    verbose=1, \n",
    "    mode='auto'\n",
    ")\n",
    "\n",
    "callbacks = [mc, es, rlr]\n",
    "\n",
    "def split_dataset(df):\n",
    "    train, test, ytrain, ytest = train_test_split(df['item_name'], df['category'], test_size=0.2, random_state=88)\n",
    "\n",
    "    return train, test, ytrain, ytest\n",
    "    \n",
    "\n",
    "def create_one_hot_labels(Y_train, Y_test, labels):\n",
    "    encoder = LabelEncoder()\n",
    "    encoder.fit(labels)\n",
    "\n",
    "    Y_train = encoder.transform(Y_train)\n",
    "    Y_test = encoder.transform(Y_test)\n",
    "\n",
    "    n_classes = np.max(Y_train) + 1\n",
    "\n",
    "    Y_train = np_utils.to_categorical(Y_train, n_classes)\n",
    "    Y_test = np_utils.to_categorical(Y_test, n_classes)\n",
    "\n",
    "    return Y_train, Y_test, n_classes, encoder\n",
    "\n",
    "\n",
    "def get_max_len(df):\n",
    "    idx = df['item_name'].str.len().idxmax()\n",
    "    print(idx)\n",
    "\n",
    "    max_len = len(df[\"item_name\"][idx])\n",
    "\n",
    "    if max_len > 450:\n",
    "        max_len = 500\n",
    "    else:\n",
    "        max_len += 50\n",
    "\n",
    "    return max_len\n",
    "\n",
    "\n",
    "def get_vectors(df, X_train, X_test, max_len):\n",
    "    vectorizer = TextVectorization(output_sequence_length=max_len)\n",
    "\n",
    "    vectorizer.adapt(df['item_name'])\n",
    "\n",
    "    vocab = vectorizer.get_vocabulary()\n",
    "\n",
    "    word_index = dict(zip(vocab, range(len(vocab))))\n",
    "\n",
    "    X_train = vectorizer(X_train)\n",
    "    X_test = vectorizer(X_test)\n",
    "\n",
    "    return X_train, X_test, vocab, word_index, vectorizer\n",
    "\n",
    "\n",
    "def get_embedding_model(matrix, n_tokens, max_len, n_classes, dim=100, lr=0.0003):\n",
    "    embedding_layer = l.Embedding(\n",
    "    n_tokens,\n",
    "    dim,\n",
    "    input_length=max_len,\n",
    "    embeddings_initializer=tf.keras.initializers.Constant(matrix),\n",
    "    trainable=True\n",
    "    )\n",
    "\n",
    "    embedding_model = Sequential()\n",
    "    embedding_model.add(embedding_layer)\n",
    "    embedding_model.add(l.GlobalMaxPooling1D()) # Turn 2D output into 1D for dense layers\n",
    "\n",
    "    embedding_model.add(l.Dense(n_classes, activation='softmax'))\n",
    "\n",
    "    optim = tf.keras.optimizers.Adam(learning_rate=lr)\n",
    "\n",
    "    embedding_model.compile(\n",
    "        loss=\"categorical_crossentropy\", optimizer=optim, metrics=['accuracy']\n",
    "    )\n",
    "\n",
    "    return embedding_model\n",
    "\n",
    "\n",
    "def get_model_acc(X_train, X_test, Y_train, Y_test, model, n_epochs=1000):\n",
    "    history = model.fit(X_train, Y_train, batch_size=25, epochs=n_epochs, validation_data=(X_test, Y_test), callbacks=callbacks)\n",
    "    metrics = [[history.history['val_accuracy'][-1], history.history['val_loss'][-1]]]\n",
    "    \n",
    "    metrics = pd.DataFrame(metrics, columns=['val_accuracy', 'val_loss'])\n",
    "    return metrics, history\n",
    "\n",
    "\n",
    "def train_ft(df):\n",
    "    cor_path = datapath(\"lee_background.cor\")\n",
    "    ft = FastText(vector_size=100)\n",
    "\n",
    "    train_ft = df['item_name'].tolist()\n",
    "\n",
    "    train_ft = [x.split() for x in train_ft]\n",
    "\n",
    "    ft.build_vocab(corpus_file=cor_path)\n",
    "\n",
    "    ft.train(\n",
    "        corpus_file=cor_path, epochs=ft.epochs,\n",
    "        total_examples=ft.corpus_count, total_words=ft.corpus_total_words,\n",
    "    )\n",
    "\n",
    "    ft.build_vocab(train_ft, update=True)\n",
    "    ft.train(train_ft, total_examples=len(train_ft), epochs=ft.epochs)\n",
    "\n",
    "    return ft\n",
    "\n",
    "\n",
    "def get_ft_matrix(vocab, word_index, ft):\n",
    "    n_tokens = len(vocab) + 2\n",
    "    embedding_dim = 100\n",
    "    hits = 0\n",
    "    misses = 0\n",
    "\n",
    "    matrix = np.zeros((n_tokens, embedding_dim))\n",
    "\n",
    "    for word, i in word_index.items():\n",
    "        try:\n",
    "            embedding_vector = ft.wv[word]\n",
    "\n",
    "            if embedding_vector is not None:\n",
    "                matrix[i] = embedding_vector\n",
    "                hits += 1\n",
    "            else:\n",
    "                matrix[i] = np.random.randn(embedding_dim)\n",
    "                misses += 1\n",
    "        except Exception as e:\n",
    "            matrix[i] = np.random.randn(embedding_dim)\n",
    "            print(e)\n",
    "            misses += 1\n",
    "\n",
    "    print(\"Converted %d words (%d misses)\" % (hits, misses))\n",
    "\n",
    "    return matrix, n_tokens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "def train_val_graph(history, metric):\n",
    "    plt.plot(history.history[metric])\n",
    "    plt.plot(history.history['val_'+metric])\n",
    "\n",
    "    plt.title(f'History of {metric} over epochs')\n",
    "    plt.xlabel(\"Epochs\")\n",
    "    plt.ylabel(metric)\n",
    "    plt.legend(['training '+metric, 'validation '+metric])\n",
    "\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_textcnn_model(matrix, n_tokens, max_len, n_classes, dim=100):\n",
    "    inputs = tf.keras.Input(shape=(max_len,))\n",
    "\n",
    "    embedding_layer = l.Embedding(\n",
    "        n_tokens,\n",
    "        dim,\n",
    "        input_length=max_len,\n",
    "        embeddings_initializer=tf.keras.initializers.Constant(matrix),\n",
    "        trainable = True\n",
    "    )(inputs)\n",
    "\n",
    "    reshape = l.Reshape((max_len, 100, 1))(embedding_layer)\n",
    "\n",
    "    conv_0 = l.Conv2D(100, kernel_size=(3,100), activation='relu')(reshape)\n",
    "    conv_1 = l.Conv2D(100, kernel_size=(4,100), activation='relu')(reshape)\n",
    "    conv_2 = l.Conv2D(100, kernel_size=(5,100), activation='relu')(reshape)\n",
    "\n",
    "    maxpool_0 = l.MaxPooling2D(pool_size=(max_len - 3 + 1, 1))(conv_0)\n",
    "    maxpool_1 = l.MaxPooling2D(pool_size=(max_len - 4 + 1, 1))(conv_1)\n",
    "    maxpool_2 = l.MaxPooling2D(pool_size=(max_len - 5 + 1, 1))(conv_2)\n",
    "\n",
    "    concatenate = l.Concatenate(axis=1)([maxpool_0, maxpool_1, maxpool_2])\n",
    "\n",
    "    flatten = l.Flatten()(concatenate)\n",
    "\n",
    "    dropout = l.Dropout(0.5)(flatten)\n",
    "\n",
    "    output = l.Dense(n_classes, activation='softmax')(dropout)\n",
    "\n",
    "    embedding_model = tf.keras.Model(inputs=inputs, outputs=output)\n",
    "\n",
    "    optim = tf.keras.optimizers.Adam(learning_rate=0.0001)\n",
    "\n",
    "    embedding_model.compile(\n",
    "        loss=\"categorical_crossentropy\", optimizer=optim, metrics=['accuracy']\n",
    "    )\n",
    "\n",
    "    return embedding_model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 414,
   "metadata": {},
   "outputs": [],
   "source": [
    "def run_textcnn(df, labels):\n",
    "    df['item_name'] = df['item_name'].astype(str)\n",
    "\n",
    "    X_train, X_test, Y_train, Y_test = split_dataset(df)\n",
    "\n",
    "    Y_train, Y_test, n_classes, encoder = create_one_hot_labels(Y_train, Y_test, labels)\n",
    "\n",
    "    max_len = get_max_len(df)\n",
    "\n",
    "    X_train, X_test, vocab, word_index, vectoriser = get_vectors(df, X_train, X_test, max_len)\n",
    "\n",
    "    ft_embeddings = train_ft(df)\n",
    "\n",
    "    ft_matrix, n_tokens = get_ft_matrix(vocab, word_index, ft_embeddings)\n",
    "\n",
    "    print(n_tokens, max_len)\n",
    "\n",
    "    ft_model = get_textcnn_model(ft_matrix, n_tokens, max_len, n_classes)\n",
    "    \n",
    "    metrics, history = get_model_acc(X_train, X_test, Y_train, Y_test, ft_model)\n",
    "    \n",
    "    return metrics, history, vectoriser, encoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 415,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "149\n",
      "Converted 611 words (0 misses)\n",
      "613 113\n",
      "Epoch 1/1000\n",
      "21/21 [==============================] - 1s 24ms/step - loss: 2.6755 - accuracy: 0.4648 - val_loss: 2.5699 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 2/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 2.5006 - accuracy: 0.5162 - val_loss: 2.3887 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 3/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 2.3237 - accuracy: 0.5162 - val_loss: 2.2266 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 4/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 2.2160 - accuracy: 0.5162 - val_loss: 2.0982 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 5/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 2.1126 - accuracy: 0.5162 - val_loss: 2.0128 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 6/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 2.0395 - accuracy: 0.5162 - val_loss: 1.9506 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 7/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.9954 - accuracy: 0.5162 - val_loss: 1.9035 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 8/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.9365 - accuracy: 0.5162 - val_loss: 1.8637 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 9/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.9183 - accuracy: 0.5162 - val_loss: 1.8285 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 10/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.9034 - accuracy: 0.5162 - val_loss: 1.7948 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 11/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.8793 - accuracy: 0.5162 - val_loss: 1.7653 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 12/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.8283 - accuracy: 0.5162 - val_loss: 1.7388 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 13/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.8149 - accuracy: 0.5162 - val_loss: 1.7116 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 14/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.7956 - accuracy: 0.5143 - val_loss: 1.6879 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 15/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.7681 - accuracy: 0.5181 - val_loss: 1.6653 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 16/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.7370 - accuracy: 0.5162 - val_loss: 1.6477 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 17/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.7420 - accuracy: 0.5181 - val_loss: 1.6287 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 18/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.7049 - accuracy: 0.5162 - val_loss: 1.6107 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 19/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.6812 - accuracy: 0.5200 - val_loss: 1.5937 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 20/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.6881 - accuracy: 0.5181 - val_loss: 1.5805 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 21/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.6754 - accuracy: 0.5162 - val_loss: 1.5715 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 22/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 1.6494 - accuracy: 0.5200 - val_loss: 1.5587 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 23/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.6260 - accuracy: 0.5200 - val_loss: 1.5476 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 24/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.6192 - accuracy: 0.5181 - val_loss: 1.5354 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 25/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.6008 - accuracy: 0.5162 - val_loss: 1.5231 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 26/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.5711 - accuracy: 0.5200 - val_loss: 1.5098 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 27/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.5847 - accuracy: 0.5219 - val_loss: 1.4985 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 28/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.5610 - accuracy: 0.5257 - val_loss: 1.4889 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 29/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.5352 - accuracy: 0.5181 - val_loss: 1.4780 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 30/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.5226 - accuracy: 0.5314 - val_loss: 1.4674 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 31/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.5195 - accuracy: 0.5238 - val_loss: 1.4544 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 32/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.4937 - accuracy: 0.5371 - val_loss: 1.4443 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 33/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.4884 - accuracy: 0.5295 - val_loss: 1.4333 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 34/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.4764 - accuracy: 0.5333 - val_loss: 1.4232 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 35/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.4491 - accuracy: 0.5410 - val_loss: 1.4114 - val_accuracy: 0.5455 - lr: 1.0000e-04\n",
      "Epoch 36/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 1.4277 - accuracy: 0.5562 - val_loss: 1.4006 - val_accuracy: 0.5530 - lr: 1.0000e-04\n",
      "Epoch 37/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 1.4089 - accuracy: 0.5467 - val_loss: 1.3892 - val_accuracy: 0.5606 - lr: 1.0000e-04\n",
      "Epoch 38/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.3970 - accuracy: 0.5600 - val_loss: 1.3786 - val_accuracy: 0.5606 - lr: 1.0000e-04\n",
      "Epoch 39/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.3689 - accuracy: 0.5810 - val_loss: 1.3638 - val_accuracy: 0.5606 - lr: 1.0000e-04\n",
      "Epoch 40/1000\n",
      "21/21 [==============================] - 0s 8ms/step - loss: 1.3515 - accuracy: 0.5886 - val_loss: 1.3507 - val_accuracy: 0.5758 - lr: 1.0000e-04\n",
      "Epoch 41/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 1.3308 - accuracy: 0.5886 - val_loss: 1.3370 - val_accuracy: 0.5833 - lr: 1.0000e-04\n",
      "Epoch 42/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 1.3045 - accuracy: 0.6114 - val_loss: 1.3250 - val_accuracy: 0.5909 - lr: 1.0000e-04\n",
      "Epoch 43/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.3017 - accuracy: 0.6057 - val_loss: 1.3145 - val_accuracy: 0.5833 - lr: 1.0000e-04\n",
      "Epoch 44/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.2585 - accuracy: 0.6133 - val_loss: 1.3013 - val_accuracy: 0.5909 - lr: 1.0000e-04\n",
      "Epoch 45/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 1.2475 - accuracy: 0.6324 - val_loss: 1.2908 - val_accuracy: 0.5985 - lr: 1.0000e-04\n",
      "Epoch 46/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.2226 - accuracy: 0.6286 - val_loss: 1.2797 - val_accuracy: 0.5985 - lr: 1.0000e-04\n",
      "Epoch 47/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 1.2209 - accuracy: 0.6229 - val_loss: 1.2667 - val_accuracy: 0.6061 - lr: 1.0000e-04\n",
      "Epoch 48/1000\n",
      "21/21 [==============================] - 0s 9ms/step - loss: 1.1952 - accuracy: 0.6457 - val_loss: 1.2569 - val_accuracy: 0.6212 - lr: 1.0000e-04\n",
      "Epoch 49/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.1794 - accuracy: 0.6476 - val_loss: 1.2454 - val_accuracy: 0.6212 - lr: 1.0000e-04\n",
      "Epoch 50/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.1612 - accuracy: 0.6495 - val_loss: 1.2351 - val_accuracy: 0.6212 - lr: 1.0000e-04\n",
      "Epoch 51/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.1460 - accuracy: 0.6705 - val_loss: 1.2237 - val_accuracy: 0.6212 - lr: 1.0000e-04\n",
      "Epoch 52/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 1.1419 - accuracy: 0.6705 - val_loss: 1.2135 - val_accuracy: 0.6288 - lr: 1.0000e-04\n",
      "Epoch 53/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 1.1191 - accuracy: 0.6667 - val_loss: 1.2031 - val_accuracy: 0.6439 - lr: 1.0000e-04\n",
      "Epoch 54/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.0927 - accuracy: 0.6876 - val_loss: 1.1933 - val_accuracy: 0.6439 - lr: 1.0000e-04\n",
      "Epoch 55/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.0926 - accuracy: 0.6914 - val_loss: 1.1850 - val_accuracy: 0.6439 - lr: 1.0000e-04\n",
      "Epoch 56/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 1.0699 - accuracy: 0.6971 - val_loss: 1.1761 - val_accuracy: 0.6591 - lr: 1.0000e-04\n",
      "Epoch 57/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 1.0496 - accuracy: 0.7029 - val_loss: 1.1648 - val_accuracy: 0.6667 - lr: 1.0000e-04\n",
      "Epoch 58/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.0234 - accuracy: 0.7048 - val_loss: 1.1561 - val_accuracy: 0.6591 - lr: 1.0000e-04\n",
      "Epoch 59/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 1.0123 - accuracy: 0.7105 - val_loss: 1.1464 - val_accuracy: 0.6667 - lr: 1.0000e-04\n",
      "Epoch 60/1000\n",
      "21/21 [==============================] - 0s 8ms/step - loss: 1.0077 - accuracy: 0.7124 - val_loss: 1.1358 - val_accuracy: 0.6742 - lr: 1.0000e-04\n",
      "Epoch 61/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 0.9846 - accuracy: 0.7352 - val_loss: 1.1275 - val_accuracy: 0.6894 - lr: 1.0000e-04\n",
      "Epoch 62/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.9865 - accuracy: 0.7371 - val_loss: 1.1210 - val_accuracy: 0.6894 - lr: 1.0000e-04\n",
      "Epoch 63/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 0.9431 - accuracy: 0.7257 - val_loss: 1.1089 - val_accuracy: 0.6970 - lr: 1.0000e-04\n",
      "Epoch 64/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.9254 - accuracy: 0.7467 - val_loss: 1.1010 - val_accuracy: 0.6970 - lr: 1.0000e-04\n",
      "Epoch 65/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.9255 - accuracy: 0.7505 - val_loss: 1.0950 - val_accuracy: 0.6970 - lr: 1.0000e-04\n",
      "Epoch 66/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 0.9340 - accuracy: 0.7448 - val_loss: 1.0859 - val_accuracy: 0.7045 - lr: 1.0000e-04\n",
      "Epoch 67/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.9050 - accuracy: 0.7486 - val_loss: 1.0798 - val_accuracy: 0.7045 - lr: 1.0000e-04\n",
      "Epoch 68/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 0.8873 - accuracy: 0.7619 - val_loss: 1.0728 - val_accuracy: 0.7348 - lr: 1.0000e-04\n",
      "Epoch 69/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.8918 - accuracy: 0.7733 - val_loss: 1.0646 - val_accuracy: 0.7348 - lr: 1.0000e-04\n",
      "Epoch 70/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.8634 - accuracy: 0.7676 - val_loss: 1.0569 - val_accuracy: 0.7348 - lr: 1.0000e-04\n",
      "Epoch 71/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.8441 - accuracy: 0.7886 - val_loss: 1.0486 - val_accuracy: 0.7348 - lr: 1.0000e-04\n",
      "Epoch 72/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 0.8511 - accuracy: 0.7771 - val_loss: 1.0391 - val_accuracy: 0.7424 - lr: 1.0000e-04\n",
      "Epoch 73/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.8086 - accuracy: 0.8057 - val_loss: 1.0314 - val_accuracy: 0.7348 - lr: 1.0000e-04\n",
      "Epoch 74/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.8104 - accuracy: 0.7867 - val_loss: 1.0240 - val_accuracy: 0.7424 - lr: 1.0000e-04\n",
      "Epoch 75/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.8112 - accuracy: 0.7962 - val_loss: 1.0196 - val_accuracy: 0.7424 - lr: 1.0000e-04\n",
      "Epoch 76/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.7889 - accuracy: 0.8095 - val_loss: 1.0113 - val_accuracy: 0.7424 - lr: 1.0000e-04\n",
      "Epoch 77/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 0.7669 - accuracy: 0.8114 - val_loss: 1.0045 - val_accuracy: 0.7500 - lr: 1.0000e-04\n",
      "Epoch 78/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.7746 - accuracy: 0.8076 - val_loss: 0.9986 - val_accuracy: 0.7500 - lr: 1.0000e-04\n",
      "Epoch 79/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.7566 - accuracy: 0.7962 - val_loss: 0.9944 - val_accuracy: 0.7500 - lr: 1.0000e-04\n",
      "Epoch 80/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.7476 - accuracy: 0.8114 - val_loss: 0.9879 - val_accuracy: 0.7500 - lr: 1.0000e-04\n",
      "Epoch 81/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.7530 - accuracy: 0.8057 - val_loss: 0.9769 - val_accuracy: 0.7500 - lr: 1.0000e-04\n",
      "Epoch 82/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.7263 - accuracy: 0.8057 - val_loss: 0.9721 - val_accuracy: 0.7500 - lr: 1.0000e-04\n",
      "Epoch 83/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.7353 - accuracy: 0.8133 - val_loss: 0.9658 - val_accuracy: 0.7500 - lr: 1.0000e-04\n",
      "Epoch 84/1000\n",
      "21/21 [==============================] - 0s 8ms/step - loss: 0.7029 - accuracy: 0.8133 - val_loss: 0.9597 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 85/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.6847 - accuracy: 0.8286 - val_loss: 0.9557 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 86/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.6960 - accuracy: 0.8267 - val_loss: 0.9501 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 87/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.6578 - accuracy: 0.8248 - val_loss: 0.9441 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 88/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.6680 - accuracy: 0.8248 - val_loss: 0.9412 - val_accuracy: 0.7576 - lr: 1.0000e-04\n",
      "Epoch 89/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.6506 - accuracy: 0.8267 - val_loss: 0.9353 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 90/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 0.6443 - accuracy: 0.8419 - val_loss: 0.9278 - val_accuracy: 0.7727 - lr: 1.0000e-04\n",
      "Epoch 91/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.6301 - accuracy: 0.8324 - val_loss: 0.9210 - val_accuracy: 0.7727 - lr: 1.0000e-04\n",
      "Epoch 92/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.6106 - accuracy: 0.8419 - val_loss: 0.9183 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 93/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.6142 - accuracy: 0.8362 - val_loss: 0.9157 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 94/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.6304 - accuracy: 0.8286 - val_loss: 0.9077 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 95/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.6226 - accuracy: 0.8286 - val_loss: 0.9053 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 96/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.6108 - accuracy: 0.8248 - val_loss: 0.8986 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 97/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.5927 - accuracy: 0.8400 - val_loss: 0.8945 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 98/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.5740 - accuracy: 0.8438 - val_loss: 0.8907 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 99/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.5885 - accuracy: 0.8324 - val_loss: 0.8864 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 100/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.5741 - accuracy: 0.8381 - val_loss: 0.8848 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 101/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.5675 - accuracy: 0.8476 - val_loss: 0.8814 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 102/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.5616 - accuracy: 0.8438 - val_loss: 0.8761 - val_accuracy: 0.7727 - lr: 1.0000e-04\n",
      "Epoch 103/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.5360 - accuracy: 0.8552 - val_loss: 0.8742 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 104/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.5503 - accuracy: 0.8571 - val_loss: 0.8686 - val_accuracy: 0.7652 - lr: 1.0000e-04\n",
      "Epoch 105/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.5319 - accuracy: 0.8495 - val_loss: 0.8675 - val_accuracy: 0.7727 - lr: 1.0000e-04\n",
      "Epoch 106/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.5115 - accuracy: 0.8514 - val_loss: 0.8626 - val_accuracy: 0.7727 - lr: 1.0000e-04\n",
      "Epoch 107/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.5300 - accuracy: 0.8457 - val_loss: 0.8599 - val_accuracy: 0.7727 - lr: 1.0000e-04\n",
      "Epoch 108/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 0.5165 - accuracy: 0.8610 - val_loss: 0.8552 - val_accuracy: 0.7803 - lr: 1.0000e-04\n",
      "Epoch 109/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.5048 - accuracy: 0.8590 - val_loss: 0.8509 - val_accuracy: 0.7803 - lr: 1.0000e-04\n",
      "Epoch 110/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.5049 - accuracy: 0.8686 - val_loss: 0.8486 - val_accuracy: 0.7803 - lr: 1.0000e-04\n",
      "Epoch 111/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.4909 - accuracy: 0.8629 - val_loss: 0.8468 - val_accuracy: 0.7803 - lr: 1.0000e-04\n",
      "Epoch 112/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.4852 - accuracy: 0.8610 - val_loss: 0.8457 - val_accuracy: 0.7803 - lr: 1.0000e-04\n",
      "Epoch 113/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.4855 - accuracy: 0.8590 - val_loss: 0.8437 - val_accuracy: 0.7803 - lr: 1.0000e-04\n",
      "Epoch 114/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 0.4650 - accuracy: 0.8724 - val_loss: 0.8412 - val_accuracy: 0.7879 - lr: 1.0000e-04\n",
      "Epoch 115/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 0.4701 - accuracy: 0.8762 - val_loss: 0.8381 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 116/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.4668 - accuracy: 0.8724 - val_loss: 0.8365 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 117/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.4606 - accuracy: 0.8819 - val_loss: 0.8360 - val_accuracy: 0.7955 - lr: 1.0000e-04\n",
      "Epoch 118/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.4540 - accuracy: 0.8743 - val_loss: 0.8365 - val_accuracy: 0.7955 - lr: 1.0000e-04\n",
      "Epoch 119/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.4494 - accuracy: 0.8743 - val_loss: 0.8347 - val_accuracy: 0.7955 - lr: 1.0000e-04\n",
      "Epoch 120/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.4198 - accuracy: 0.8933 - val_loss: 0.8315 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 121/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.4440 - accuracy: 0.8838 - val_loss: 0.8292 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 122/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.4277 - accuracy: 0.8819 - val_loss: 0.8326 - val_accuracy: 0.7955 - lr: 1.0000e-04\n",
      "Epoch 123/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.4201 - accuracy: 0.8819 - val_loss: 0.8287 - val_accuracy: 0.7955 - lr: 1.0000e-04\n",
      "Epoch 124/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.4408 - accuracy: 0.8667 - val_loss: 0.8267 - val_accuracy: 0.7955 - lr: 1.0000e-04\n",
      "Epoch 125/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.4255 - accuracy: 0.8838 - val_loss: 0.8232 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 126/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.4308 - accuracy: 0.8914 - val_loss: 0.8222 - val_accuracy: 0.7955 - lr: 1.0000e-04\n",
      "Epoch 127/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.4194 - accuracy: 0.8876 - val_loss: 0.8235 - val_accuracy: 0.7955 - lr: 1.0000e-04\n",
      "Epoch 128/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.4243 - accuracy: 0.8895 - val_loss: 0.8203 - val_accuracy: 0.7955 - lr: 1.0000e-04\n",
      "Epoch 129/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.4001 - accuracy: 0.8857 - val_loss: 0.8211 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 130/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3962 - accuracy: 0.8952 - val_loss: 0.8180 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 131/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3993 - accuracy: 0.8838 - val_loss: 0.8183 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 132/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.4032 - accuracy: 0.8990 - val_loss: 0.8179 - val_accuracy: 0.7955 - lr: 1.0000e-04\n",
      "Epoch 133/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3983 - accuracy: 0.8971 - val_loss: 0.8174 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 134/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3813 - accuracy: 0.8990 - val_loss: 0.8158 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 135/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3906 - accuracy: 0.8971 - val_loss: 0.8149 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 136/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3788 - accuracy: 0.9010 - val_loss: 0.8192 - val_accuracy: 0.7955 - lr: 1.0000e-04\n",
      "Epoch 137/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3706 - accuracy: 0.8952 - val_loss: 0.8182 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 138/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3819 - accuracy: 0.8914 - val_loss: 0.8183 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 139/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3554 - accuracy: 0.9048 - val_loss: 0.8208 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 140/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.3626 - accuracy: 0.9067 - val_loss: 0.8173 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 141/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3579 - accuracy: 0.9105 - val_loss: 0.8161 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 142/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3621 - accuracy: 0.9029 - val_loss: 0.8169 - val_accuracy: 0.7955 - lr: 1.0000e-04\n",
      "Epoch 143/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3655 - accuracy: 0.9086 - val_loss: 0.8173 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 144/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3457 - accuracy: 0.9124 - val_loss: 0.8171 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 145/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3239 - accuracy: 0.9162 - val_loss: 0.8168 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 146/1000\n",
      "21/21 [==============================] - 0s 7ms/step - loss: 0.3423 - accuracy: 0.9124 - val_loss: 0.8118 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 147/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3286 - accuracy: 0.9200 - val_loss: 0.8119 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 148/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3321 - accuracy: 0.9181 - val_loss: 0.8121 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 149/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3261 - accuracy: 0.9238 - val_loss: 0.8135 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 150/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3205 - accuracy: 0.9276 - val_loss: 0.8135 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 151/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.3286 - accuracy: 0.9162 - val_loss: 0.8144 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 152/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3337 - accuracy: 0.9200 - val_loss: 0.8118 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 153/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3211 - accuracy: 0.9200 - val_loss: 0.8119 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 154/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.3188 - accuracy: 0.9181 - val_loss: 0.8124 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 155/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3226 - accuracy: 0.9086 - val_loss: 0.8142 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 156/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3199 - accuracy: 0.9124 - val_loss: 0.8155 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 157/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2989 - accuracy: 0.9314 - val_loss: 0.8113 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 158/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3040 - accuracy: 0.9257 - val_loss: 0.8139 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 159/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3002 - accuracy: 0.9314 - val_loss: 0.8100 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 160/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2866 - accuracy: 0.9276 - val_loss: 0.8123 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 161/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.3012 - accuracy: 0.9143 - val_loss: 0.8126 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 162/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.2785 - accuracy: 0.9390 - val_loss: 0.8137 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 163/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.2921 - accuracy: 0.9314 - val_loss: 0.8065 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 164/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2835 - accuracy: 0.9333 - val_loss: 0.8045 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 165/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2792 - accuracy: 0.9257 - val_loss: 0.8100 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 166/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2748 - accuracy: 0.9295 - val_loss: 0.8130 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 167/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2781 - accuracy: 0.9352 - val_loss: 0.8117 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 168/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2657 - accuracy: 0.9390 - val_loss: 0.8115 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 169/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2650 - accuracy: 0.9352 - val_loss: 0.8131 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 170/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2560 - accuracy: 0.9467 - val_loss: 0.8125 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 171/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.2756 - accuracy: 0.9333 - val_loss: 0.8112 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 172/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.2545 - accuracy: 0.9448 - val_loss: 0.8060 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 173/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2633 - accuracy: 0.9448 - val_loss: 0.8072 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 174/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.2545 - accuracy: 0.9333 - val_loss: 0.8116 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 175/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.2696 - accuracy: 0.9276 - val_loss: 0.8091 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 176/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2443 - accuracy: 0.9486 - val_loss: 0.8092 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 177/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2292 - accuracy: 0.9467 - val_loss: 0.8097 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 178/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2500 - accuracy: 0.9467 - val_loss: 0.8098 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 179/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2342 - accuracy: 0.9524 - val_loss: 0.8125 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 180/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2305 - accuracy: 0.9448 - val_loss: 0.8165 - val_accuracy: 0.7955 - lr: 1.0000e-04\n",
      "Epoch 181/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2396 - accuracy: 0.9390 - val_loss: 0.8123 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 182/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2218 - accuracy: 0.9505 - val_loss: 0.8106 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 183/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2333 - accuracy: 0.9543 - val_loss: 0.8114 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 184/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.2265 - accuracy: 0.9448 - val_loss: 0.8152 - val_accuracy: 0.7955 - lr: 1.0000e-04\n",
      "Epoch 185/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2144 - accuracy: 0.9524 - val_loss: 0.8144 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 186/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.2238 - accuracy: 0.9543 - val_loss: 0.8128 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 187/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.2244 - accuracy: 0.9505 - val_loss: 0.8152 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 188/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.2240 - accuracy: 0.9524 - val_loss: 0.8133 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 189/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.2258 - accuracy: 0.9543 - val_loss: 0.8139 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 190/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2009 - accuracy: 0.9562 - val_loss: 0.8156 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 191/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.2255 - accuracy: 0.9524 - val_loss: 0.8142 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 192/1000\n",
      "21/21 [==============================] - 0s 8ms/step - loss: 0.2178 - accuracy: 0.9524 - val_loss: 0.8136 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 193/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1980 - accuracy: 0.9619 - val_loss: 0.8167 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 194/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.2063 - accuracy: 0.9524 - val_loss: 0.8176 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 195/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1978 - accuracy: 0.9581 - val_loss: 0.8150 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 196/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1982 - accuracy: 0.9543 - val_loss: 0.8110 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 197/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1967 - accuracy: 0.9581 - val_loss: 0.8135 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 198/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.1937 - accuracy: 0.9524 - val_loss: 0.8190 - val_accuracy: 0.8030 - lr: 1.0000e-04\n",
      "Epoch 199/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1915 - accuracy: 0.9581 - val_loss: 0.8168 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 200/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1948 - accuracy: 0.9562 - val_loss: 0.8147 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 201/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.2003 - accuracy: 0.9581 - val_loss: 0.8154 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 202/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1984 - accuracy: 0.9600 - val_loss: 0.8169 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 203/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1945 - accuracy: 0.9524 - val_loss: 0.8143 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 204/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1753 - accuracy: 0.9619 - val_loss: 0.8130 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 205/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1697 - accuracy: 0.9600 - val_loss: 0.8089 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 206/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1888 - accuracy: 0.9543 - val_loss: 0.8158 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 207/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1793 - accuracy: 0.9638 - val_loss: 0.8176 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 208/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1702 - accuracy: 0.9676 - val_loss: 0.8156 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 209/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1692 - accuracy: 0.9619 - val_loss: 0.8158 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 210/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1732 - accuracy: 0.9619 - val_loss: 0.8207 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 211/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1623 - accuracy: 0.9676 - val_loss: 0.8177 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 212/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1665 - accuracy: 0.9638 - val_loss: 0.8160 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 213/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1748 - accuracy: 0.9619 - val_loss: 0.8164 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 214/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1709 - accuracy: 0.9600 - val_loss: 0.8194 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 215/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1590 - accuracy: 0.9638 - val_loss: 0.8185 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 216/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1636 - accuracy: 0.9676 - val_loss: 0.8131 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 217/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1454 - accuracy: 0.9676 - val_loss: 0.8133 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 218/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1568 - accuracy: 0.9657 - val_loss: 0.8166 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 219/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1614 - accuracy: 0.9638 - val_loss: 0.8190 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 220/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1575 - accuracy: 0.9581 - val_loss: 0.8193 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 221/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1425 - accuracy: 0.9714 - val_loss: 0.8231 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 222/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1488 - accuracy: 0.9676 - val_loss: 0.8211 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 223/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1398 - accuracy: 0.9752 - val_loss: 0.8189 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 224/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1447 - accuracy: 0.9695 - val_loss: 0.8192 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 225/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1408 - accuracy: 0.9695 - val_loss: 0.8161 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 226/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1460 - accuracy: 0.9638 - val_loss: 0.8161 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 227/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1403 - accuracy: 0.9695 - val_loss: 0.8172 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 228/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1341 - accuracy: 0.9733 - val_loss: 0.8197 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 229/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1394 - accuracy: 0.9676 - val_loss: 0.8207 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 230/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.1363 - accuracy: 0.9752 - val_loss: 0.8177 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 231/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1304 - accuracy: 0.9771 - val_loss: 0.8166 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 232/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.1277 - accuracy: 0.9733 - val_loss: 0.8179 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 233/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1318 - accuracy: 0.9695 - val_loss: 0.8207 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 234/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1277 - accuracy: 0.9790 - val_loss: 0.8211 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 235/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1302 - accuracy: 0.9714 - val_loss: 0.8204 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 236/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1251 - accuracy: 0.9733 - val_loss: 0.8222 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 237/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1224 - accuracy: 0.9810 - val_loss: 0.8242 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 238/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1205 - accuracy: 0.9810 - val_loss: 0.8244 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 239/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1121 - accuracy: 0.9790 - val_loss: 0.8253 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 240/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1241 - accuracy: 0.9752 - val_loss: 0.8250 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 241/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.1157 - accuracy: 0.9733 - val_loss: 0.8260 - val_accuracy: 0.8182 - lr: 1.0000e-04\n",
      "Epoch 242/1000\n",
      "16/21 [=====================>........] - ETA: 0s - loss: 0.1259 - accuracy: 0.9725\n",
      "Epoch 242: ReduceLROnPlateau reducing learning rate to 7.999999797903001e-05.\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.1193 - accuracy: 0.9752 - val_loss: 0.8243 - val_accuracy: 0.8106 - lr: 1.0000e-04\n",
      "Epoch 243/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.1168 - accuracy: 0.9790 - val_loss: 0.8231 - val_accuracy: 0.8106 - lr: 8.0000e-05\n",
      "Epoch 244/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.1144 - accuracy: 0.9771 - val_loss: 0.8229 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 245/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1266 - accuracy: 0.9733 - val_loss: 0.8232 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 246/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.1178 - accuracy: 0.9790 - val_loss: 0.8232 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 247/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.1208 - accuracy: 0.9771 - val_loss: 0.8232 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 248/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.1044 - accuracy: 0.9848 - val_loss: 0.8231 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 249/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1042 - accuracy: 0.9790 - val_loss: 0.8248 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 250/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1113 - accuracy: 0.9733 - val_loss: 0.8247 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 251/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1061 - accuracy: 0.9810 - val_loss: 0.8262 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 252/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1113 - accuracy: 0.9714 - val_loss: 0.8243 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 253/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0989 - accuracy: 0.9771 - val_loss: 0.8233 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 254/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1127 - accuracy: 0.9771 - val_loss: 0.8219 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 255/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1045 - accuracy: 0.9810 - val_loss: 0.8266 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 256/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1091 - accuracy: 0.9752 - val_loss: 0.8266 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 257/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0987 - accuracy: 0.9771 - val_loss: 0.8264 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 258/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0976 - accuracy: 0.9848 - val_loss: 0.8227 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 259/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0991 - accuracy: 0.9867 - val_loss: 0.8221 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 260/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1016 - accuracy: 0.9790 - val_loss: 0.8205 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 261/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0888 - accuracy: 0.9867 - val_loss: 0.8204 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 262/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0970 - accuracy: 0.9771 - val_loss: 0.8231 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 263/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0990 - accuracy: 0.9829 - val_loss: 0.8290 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 264/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0960 - accuracy: 0.9829 - val_loss: 0.8288 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 265/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0925 - accuracy: 0.9810 - val_loss: 0.8286 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 266/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0954 - accuracy: 0.9810 - val_loss: 0.8279 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 267/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.1005 - accuracy: 0.9752 - val_loss: 0.8242 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 268/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.0871 - accuracy: 0.9829 - val_loss: 0.8270 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 269/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0808 - accuracy: 0.9867 - val_loss: 0.8244 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 270/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0911 - accuracy: 0.9829 - val_loss: 0.8262 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 271/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0865 - accuracy: 0.9867 - val_loss: 0.8261 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 272/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0865 - accuracy: 0.9829 - val_loss: 0.8257 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 273/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0820 - accuracy: 0.9829 - val_loss: 0.8304 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 274/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.0734 - accuracy: 0.9924 - val_loss: 0.8296 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 275/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0772 - accuracy: 0.9886 - val_loss: 0.8324 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 276/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0815 - accuracy: 0.9867 - val_loss: 0.8350 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 277/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0870 - accuracy: 0.9848 - val_loss: 0.8334 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 278/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.0800 - accuracy: 0.9848 - val_loss: 0.8337 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 279/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.0748 - accuracy: 0.9848 - val_loss: 0.8320 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 280/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.0882 - accuracy: 0.9829 - val_loss: 0.8352 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 281/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0863 - accuracy: 0.9771 - val_loss: 0.8333 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 282/1000\n",
      "21/21 [==============================] - 0s 5ms/step - loss: 0.0753 - accuracy: 0.9848 - val_loss: 0.8337 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 283/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0818 - accuracy: 0.9790 - val_loss: 0.8309 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 284/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0781 - accuracy: 0.9905 - val_loss: 0.8336 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 285/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0842 - accuracy: 0.9848 - val_loss: 0.8339 - val_accuracy: 0.8106 - lr: 8.0000e-05\n",
      "Epoch 286/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0961 - accuracy: 0.9810 - val_loss: 0.8359 - val_accuracy: 0.8106 - lr: 8.0000e-05\n",
      "Epoch 287/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0812 - accuracy: 0.9848 - val_loss: 0.8376 - val_accuracy: 0.8106 - lr: 8.0000e-05\n",
      "Epoch 288/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0695 - accuracy: 0.9886 - val_loss: 0.8420 - val_accuracy: 0.8106 - lr: 8.0000e-05\n",
      "Epoch 289/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0691 - accuracy: 0.9886 - val_loss: 0.8381 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 290/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0811 - accuracy: 0.9848 - val_loss: 0.8386 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 291/1000\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0662 - accuracy: 0.9924 - val_loss: 0.8386 - val_accuracy: 0.8182 - lr: 8.0000e-05\n",
      "Epoch 292/1000\n",
      "18/21 [========================>.....] - ETA: 0s - loss: 0.0818 - accuracy: 0.9867\n",
      "Epoch 292: ReduceLROnPlateau reducing learning rate to 6.399999838322402e-05.\n",
      "21/21 [==============================] - 0s 4ms/step - loss: 0.0790 - accuracy: 0.9867 - val_loss: 0.8436 - val_accuracy: 0.8182 - lr: 8.0000e-05\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "metrics, history, vectoriser, encoder = run_textcnn(custom_data, custom_labels)\n",
    "\n",
    "train_val_graph(history, \"accuracy\")\n",
    "train_val_graph(history, \"loss\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 419,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_3\"\n",
      "__________________________________________________________________________________________________\n",
      " Layer (type)                   Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      " input_13 (InputLayer)          [(None, 113)]        0           []                               \n",
      "                                                                                                  \n",
      " embedding_2 (Embedding)        (None, 113, 100)     61300       ['input_13[0][0]']               \n",
      "                                                                                                  \n",
      " reshape_2 (Reshape)            (None, 113, 100, 1)  0           ['embedding_2[0][0]']            \n",
      "                                                                                                  \n",
      " conv2d_6 (Conv2D)              (None, 111, 1, 100)  30100       ['reshape_2[0][0]']              \n",
      "                                                                                                  \n",
      " conv2d_7 (Conv2D)              (None, 110, 1, 100)  40100       ['reshape_2[0][0]']              \n",
      "                                                                                                  \n",
      " conv2d_8 (Conv2D)              (None, 109, 1, 100)  50100       ['reshape_2[0][0]']              \n",
      "                                                                                                  \n",
      " max_pooling2d_6 (MaxPooling2D)  (None, 1, 1, 100)   0           ['conv2d_6[0][0]']               \n",
      "                                                                                                  \n",
      " max_pooling2d_7 (MaxPooling2D)  (None, 1, 1, 100)   0           ['conv2d_7[0][0]']               \n",
      "                                                                                                  \n",
      " max_pooling2d_8 (MaxPooling2D)  (None, 1, 1, 100)   0           ['conv2d_8[0][0]']               \n",
      "                                                                                                  \n",
      " concatenate_2 (Concatenate)    (None, 3, 1, 100)    0           ['max_pooling2d_6[0][0]',        \n",
      "                                                                  'max_pooling2d_7[0][0]',        \n",
      "                                                                  'max_pooling2d_8[0][0]']        \n",
      "                                                                                                  \n",
      " flatten_2 (Flatten)            (None, 300)          0           ['concatenate_2[0][0]']          \n",
      "                                                                                                  \n",
      " dropout_2 (Dropout)            (None, 300)          0           ['flatten_2[0][0]']              \n",
      "                                                                                                  \n",
      " dense_2 (Dense)                (None, 16)           4816        ['dropout_2[0][0]']              \n",
      "                                                                                                  \n",
      "==================================================================================================\n",
      "Total params: 186,416\n",
      "Trainable params: 186,416\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n",
      "fried chicken is labelled with the category \"groceries\" with 91.96% certainty\n",
      "poems is labelled with the category \"groceries\" with 20.18% certainty\n",
      "biscuits is labelled with the category \"groceries\" with 20.18% certainty\n",
      "ice cream is labelled with the category \"snacks\" with 99.95% certainty\n",
      "vodka is labelled with the category \"groceries\" with 20.18% certainty\n",
      "lemon tea is labelled with the category \"drinks\" with 93.04% certainty\n",
      "lemon iced tea is labelled with the category \"drinks\" with 26.37% certainty\n",
      "lemon ice tea is labelled with the category \"drinks\" with 85.25% certainty\n"
     ]
    }
   ],
   "source": [
    "model = load_model('final.h5')\n",
    "\n",
    "model.summary()\n",
    "\n",
    "test_string = tf.keras.Input(shape=(1,), dtype=\"string\")\n",
    "\n",
    "x = vectoriser(test_string)\n",
    "\n",
    "class_names = encoder.classes_\n",
    "\n",
    "preds = model(x)\n",
    "\n",
    "test_model = tf.keras.Model(test_string, preds)\n",
    "\n",
    "test_items = [['fried chicken'], ['poems'], ['biscuits'], ['ice cream'], ['vodka'], ['lemon tea'], ['lemon iced tea'], ['lemon ice tea']]\n",
    "\n",
    "probabilities = test_model.predict(test_items)\n",
    "\n",
    "for i in tf.range(len(probabilities)):\n",
    "    print(test_items[i][0] + \" is labelled with the category \\\"\" + class_names[np.argmax(probabilities[i])] + \"\\\" with \" + str(float('{0:.2f}'.format(np.max(probabilities[i])*100))) + \"% certainty\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "custom_stores = custom_data.store_name.unique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['tesco', 'lidl', 'sainsbury s', 'aldi', 'waitrose   partners',\n",
       "       't k maxx', 'boots uk limited', 'sushi jun', 'whlung supermarket',\n",
       "       'uniqlo', 'uber eats', 'taco bell', 'vapouriz', 'oxfam',\n",
       "       'blue bear bookshop', 'evapo', 'poundland', 'aldi stores',\n",
       "       'simply fresh', 't4', 'b m', 'kenji', 'www itsu com', 'greggs',\n",
       "       'tea terrace', 'osprey london', 't a shake shack', 'costa',\n",
       "       'house of fraser', 'currys pc world', 'co op', 'bodycare'],\n",
       "      dtype=object)"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "custom_stores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_stores(X2_train, X2_test, stores):\n",
    "    encoder = LabelEncoder()\n",
    "    encoder.fit(stores)\n",
    "\n",
    "    X2_train = encoder.transform(X2_train)\n",
    "    X2_test = encoder.transform(X2_test)\n",
    "\n",
    "    n_stores = np.max(X2_train) + 1\n",
    "\n",
    "    X2_train = np_utils.to_categorical(X2_train, n_stores)\n",
    "    X2_test = np_utils.to_categorical(X2_test, n_stores)\n",
    "\n",
    "    stores_shape = len(X2_train[0])\n",
    "\n",
    "    return X2_train, X2_test, stores_shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "def split_dataset2(df):\n",
    "    train, test, x2train, x2test, ytrain, ytest = train_test_split(df['item_name'].astype(str), df['store_name'], df['category'], test_size=0.2, random_state=88)\n",
    "\n",
    "    return train, test, x2train, x2test, ytrain, ytest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_store_vectors(df, X2_train, X2_test, max_len):\n",
    "    vectorizer = TextVectorization(output_sequence_length=max_len)\n",
    "\n",
    "    vectorizer.adapt(df['store_name'])\n",
    "\n",
    "    vocab = vectorizer.get_vocabulary()\n",
    "\n",
    "    word_index = dict(zip(vocab, range(len(vocab))))\n",
    "\n",
    "    X2_train = vectorizer(X2_train)\n",
    "    X2_test = vectorizer(X2_test)\n",
    "\n",
    "    print(len(X2_train[0]))\n",
    "\n",
    "    return X2_train, X2_test, vocab, word_index, vectorizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_textcnn2_model(matrix, n_tokens, store_tokens, max_len, n_classes, dim=100):\n",
    "    product_inputs = tf.keras.Input(shape=(max_len,))\n",
    "    \n",
    "    product_embedding = l.Embedding(\n",
    "        n_tokens,\n",
    "        dim,\n",
    "        input_length=max_len,\n",
    "        embeddings_initializer=tf.keras.initializers.Constant(matrix),\n",
    "        trainable = True\n",
    "    )(product_inputs)\n",
    "    \n",
    "    store_inputs = tf.keras.Input(shape=(50,))\n",
    "\n",
    "    store_embedding = l.Embedding(\n",
    "        store_tokens,\n",
    "        dim,\n",
    "        input_length=50,\n",
    "        trainable = True\n",
    "    )(store_inputs)\n",
    "\n",
    "    reshape = l.Reshape((max_len, 100, 1))(product_embedding)\n",
    "    pooled = l.GlobalMaxPooling1D()(store_embedding)\n",
    "\n",
    "    conv_0 = l.Conv2D(100, kernel_size=(3,100), activation='relu')(reshape)\n",
    "    conv_1 = l.Conv2D(100, kernel_size=(4,100), activation='relu')(reshape)\n",
    "    conv_2 = l.Conv2D(100, kernel_size=(5,100), activation='relu')(reshape)\n",
    "\n",
    "    maxpool_0 = l.MaxPooling2D(pool_size=(max_len - 3 + 1, 1))(conv_0)\n",
    "    maxpool_1 = l.MaxPooling2D(pool_size=(max_len - 4 + 1, 1))(conv_1)\n",
    "    maxpool_2 = l.MaxPooling2D(pool_size=(max_len - 5 + 1, 1))(conv_2)\n",
    "\n",
    "    concatenate = l.Concatenate(axis=1)([maxpool_0, maxpool_1, maxpool_2])\n",
    "\n",
    "    flatten = l.Flatten()(concatenate)\n",
    "\n",
    "    concatenate = l.Concatenate(axis=1)([flatten,pooled])\n",
    "\n",
    "    dropout = l.Dropout(0.5)(concatenate)\n",
    "\n",
    "    dense1 = l.Dense(128, activation='relu')(dropout)\n",
    "\n",
    "    dropout2 = l.Dropout(0.5)(dense1)\n",
    "\n",
    "    output = l.Dense(n_classes, activation='softmax')(dropout2)\n",
    "\n",
    "    embedding_model = tf.keras.Model(inputs=[product_inputs, store_inputs], outputs=output)\n",
    "\n",
    "    optim = tf.keras.optimizers.Adam(learning_rate=0.0001)\n",
    "\n",
    "    embedding_model.compile(\n",
    "        loss=\"categorical_crossentropy\", optimizer=optim, metrics=['accuracy']\n",
    "    )\n",
    "\n",
    "    return embedding_model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_model_acc2(X_train, X_test, X2_train, X2_test, Y_train, Y_test, model, n_epochs=1000):\n",
    "    history = model.fit([X_train, X2_train],Y_train, batch_size=25, epochs=n_epochs, validation_data=([X_test, X2_test], Y_test), callbacks=callbacks)\n",
    "    metrics = [[history.history['val_accuracy'][-1], history.history['val_loss'][-1]]]\n",
    "    \n",
    "    metrics = pd.DataFrame(metrics, columns=['val_accuracy', 'val_loss'])\n",
    "    return metrics, history"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "def run_storedata(df, labels):\n",
    "    df['item_name'] = df['item_name'].astype(str)\n",
    "\n",
    "    X_train, X_test, X2_train, X2_test, Y_train, Y_test = split_dataset2(df)\n",
    "\n",
    "    Y_train, Y_test, n_classes, encoder = create_one_hot_labels(Y_train, Y_test, labels)\n",
    "\n",
    "    max_len = get_max_len(df)\n",
    "    print(max_len)\n",
    "\n",
    "    X_train, X_test, vocab, word_index, vectoriser = get_vectors(df, X_train, X_test, max_len)\n",
    "    X2_train, X2_test, store_vocab, _, store_vectoriser = get_store_vectors(df, X2_train, X2_test, max_len=50)\n",
    "\n",
    "    ft_embeddings = train_ft(df)\n",
    "\n",
    "    ft_matrix, n_tokens = get_ft_matrix(vocab, word_index, ft_embeddings)\n",
    "\n",
    "    store_tokens = len(store_vocab)\n",
    "\n",
    "    print(n_tokens, max_len)\n",
    "\n",
    "    model = get_textcnn2_model(ft_matrix, n_tokens, store_tokens, max_len, n_classes)\n",
    "    \n",
    "    metrics, history = get_model_acc2(X_train, X_test, X2_train, X2_test, Y_train, Y_test, model)\n",
    "    \n",
    "    return metrics, history, model, encoder, vectoriser, store_vectoriser"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'run_storedata' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32md:\\University\\FYP\\final\\final_classification.ipynb Cell 37'\u001b[0m in \u001b[0;36m<cell line: 1>\u001b[1;34m()\u001b[0m\n\u001b[1;32m----> <a href='vscode-notebook-cell:/d%3A/University/FYP/final/final_classification.ipynb#ch0000038?line=0'>1</a>\u001b[0m metrics, history, model, encoder, vectoriser, store_vectoriser \u001b[39m=\u001b[39m run_storedata(custom_data, custom_labels)\n\u001b[0;32m      <a href='vscode-notebook-cell:/d%3A/University/FYP/final/final_classification.ipynb#ch0000038?line=2'>3</a>\u001b[0m train_val_graph(history, \u001b[39m\"\u001b[39m\u001b[39maccuracy\u001b[39m\u001b[39m\"\u001b[39m)\n\u001b[0;32m      <a href='vscode-notebook-cell:/d%3A/University/FYP/final/final_classification.ipynb#ch0000038?line=3'>4</a>\u001b[0m train_val_graph(history, \u001b[39m\"\u001b[39m\u001b[39mloss\u001b[39m\u001b[39m\"\u001b[39m)\n",
      "\u001b[1;31mNameError\u001b[0m: name 'run_storedata' is not defined"
     ]
    }
   ],
   "source": [
    "metrics, history, model, encoder, vectoriser, store_vectoriser = run_storedata(custom_data, custom_labels)\n",
    "\n",
    "train_val_graph(history, \"accuracy\")\n",
    "train_val_graph(history, \"loss\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "outfile = open('encoder', 'wb')\n",
    "\n",
    "pickle.dump(encoder, outfile)\n",
    "\n",
    "outfile.close"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "outfile = open('vectoriser', 'wb')\n",
    "\n",
    "pickle.dump(vectoriser, outfile)\n",
    "\n",
    "outfile.close"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras import backend as K\n",
    "K.clear_session()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import load_model\n",
    "K.clear_session()\n",
    "model = load_model('final.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_1\"\n",
      "__________________________________________________________________________________________________\n",
      " Layer (type)                   Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      " input_3 (InputLayer)           [(None, 120)]        0           []                               \n",
      "                                                                                                  \n",
      " embedding_2 (Embedding)        (None, 120, 100)     67100       ['input_3[0][0]']                \n",
      "                                                                                                  \n",
      " reshape_1 (Reshape)            (None, 120, 100, 1)  0           ['embedding_2[0][0]']            \n",
      "                                                                                                  \n",
      " conv2d_3 (Conv2D)              (None, 118, 1, 100)  30100       ['reshape_1[0][0]']              \n",
      "                                                                                                  \n",
      " conv2d_4 (Conv2D)              (None, 117, 1, 100)  40100       ['reshape_1[0][0]']              \n",
      "                                                                                                  \n",
      " conv2d_5 (Conv2D)              (None, 116, 1, 100)  50100       ['reshape_1[0][0]']              \n",
      "                                                                                                  \n",
      " max_pooling2d_3 (MaxPooling2D)  (None, 1, 1, 100)   0           ['conv2d_3[0][0]']               \n",
      "                                                                                                  \n",
      " max_pooling2d_4 (MaxPooling2D)  (None, 1, 1, 100)   0           ['conv2d_4[0][0]']               \n",
      "                                                                                                  \n",
      " max_pooling2d_5 (MaxPooling2D)  (None, 1, 1, 100)   0           ['conv2d_5[0][0]']               \n",
      "                                                                                                  \n",
      " input_4 (InputLayer)           [(None, 50)]         0           []                               \n",
      "                                                                                                  \n",
      " concatenate_2 (Concatenate)    (None, 3, 1, 100)    0           ['max_pooling2d_3[0][0]',        \n",
      "                                                                  'max_pooling2d_4[0][0]',        \n",
      "                                                                  'max_pooling2d_5[0][0]']        \n",
      "                                                                                                  \n",
      " embedding_3 (Embedding)        (None, 50, 100)      5900        ['input_4[0][0]']                \n",
      "                                                                                                  \n",
      " flatten_1 (Flatten)            (None, 300)          0           ['concatenate_2[0][0]']          \n",
      "                                                                                                  \n",
      " global_max_pooling1d_1 (Global  (None, 100)         0           ['embedding_3[0][0]']            \n",
      " MaxPooling1D)                                                                                    \n",
      "                                                                                                  \n",
      " concatenate_3 (Concatenate)    (None, 400)          0           ['flatten_1[0][0]',              \n",
      "                                                                  'global_max_pooling1d_1[0][0]'] \n",
      "                                                                                                  \n",
      " dropout_2 (Dropout)            (None, 400)          0           ['concatenate_3[0][0]']          \n",
      "                                                                                                  \n",
      " dense_2 (Dense)                (None, 128)          51328       ['dropout_2[0][0]']              \n",
      "                                                                                                  \n",
      " dropout_3 (Dropout)            (None, 128)          0           ['dense_2[0][0]']                \n",
      "                                                                                                  \n",
      " dense_3 (Dense)                (None, 16)           2064        ['dropout_3[0][0]']              \n",
      "                                                                                                  \n",
      "==================================================================================================\n",
      "Total params: 246,692\n",
      "Trainable params: 246,692\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf.keras.utils.plot_model(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "items = [\n",
    "    ['fried chicken'],\n",
    "    ['poems'],\n",
    "    ['biscuits'],\n",
    "    ['ice cream'],\n",
    "    ['vodka'],\n",
    "    ['lemon tea'],\n",
    "    ['lemon iced tea'],\n",
    "    ['lemon ice tea'],\n",
    "    ['fried chicken'],\n",
    "    ['poems'],\n",
    "    ['biscuits'],\n",
    "    ['ice cream'],\n",
    "    ['vodka'],\n",
    "    ['lemon tea'],\n",
    "    ['lemon iced tea'],\n",
    "    ['lemon ice tea']\n",
    "    ]\n",
    "    \n",
    "stores = [\n",
    "    ['t4'],\n",
    "    ['book shop'],\n",
    "    ['tesco'],\n",
    "    ['tesco'],\n",
    "    ['sainsburys'],\n",
    "    ['aldi'],\n",
    "    ['lidl'],\n",
    "    ['t4'],\n",
    "    ['tesco'],\n",
    "    ['tesco'],\n",
    "    ['sainsburys'],\n",
    "    ['asda'],\n",
    "    ['tesco'],\n",
    "    ['tesco'],\n",
    "    ['tesco'],\n",
    "    ['tesco']\n",
    "    ]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "'fried chicken' from 't4' is labelled with the category \"eating out\" with 80.56% certainty\n",
      "'poems' from 'book shop' is labelled with the category \"snacks\" with 20.78% certainty\n",
      "'biscuits' from 'tesco' is labelled with the category \"snacks\" with 98.2% certainty\n",
      "'ice cream' from 'tesco' is labelled with the category \"snacks\" with 99.98% certainty\n",
      "'vodka' from 'sainsburys' is labelled with the category \"snacks\" with 20.78% certainty\n",
      "'lemon tea' from 'aldi' is labelled with the category \"drinks\" with 99.64% certainty\n",
      "'lemon iced tea' from 'lidl' is labelled with the category \"drinks\" with 99.93% certainty\n",
      "'lemon ice tea' from 't4' is labelled with the category \"drinks\" with 96.87% certainty\n",
      "'fried chicken' from 'tesco' is labelled with the category \"groceries\" with 93.39% certainty\n",
      "'poems' from 'tesco' is labelled with the category \"snacks\" with 62.55% certainty\n",
      "'biscuits' from 'sainsburys' is labelled with the category \"snacks\" with 91.87% certainty\n",
      "'ice cream' from 'asda' is labelled with the category \"snacks\" with 99.92% certainty\n",
      "'vodka' from 'tesco' is labelled with the category \"snacks\" with 62.55% certainty\n",
      "'lemon tea' from 'tesco' is labelled with the category \"drinks\" with 99.69% certainty\n",
      "'lemon iced tea' from 'tesco' is labelled with the category \"drinks\" with 99.91% certainty\n",
      "'lemon ice tea' from 'tesco' is labelled with the category \"drinks\" with 94.59% certainty\n"
     ]
    }
   ],
   "source": [
    "product_string = tf.keras.Input(shape=(1,), dtype=\"string\")\n",
    "store_string = tf.keras.Input(shape=(1,), dtype=\"string\")\n",
    "\n",
    "x = vectoriser(product_string)\n",
    "x2 = store_vectoriser(store_string)\n",
    "\n",
    "class_names= encoder.classes_\n",
    "\n",
    "predictions = model([x, x2])\n",
    "\n",
    "\n",
    "e2e_model = tf.keras.Model([product_string, store_string], predictions)\n",
    "\n",
    "#e2e_model.summary()\n",
    "\n",
    "products_input = tf.convert_to_tensor(items)\n",
    "store_input = tf.convert_to_tensor(stores)\n",
    "\n",
    "\n",
    "probabilities = e2e_model.predict([products_input, store_input])\n",
    "\n",
    "for i in tf.range(len(probabilities)):\n",
    "    print(\"'\"+products_input.numpy()[i][0].decode('UTF-8') + \"' from '\" + store_input.numpy()[i][0].decode('UTF-8') + \"' is labelled with the category \\\"\" + class_names[np.argmax(probabilities[i])] + \"\\\" with \" + str(float('{0:.2f}'.format(np.max(probabilities[i])*100))) + \"% certainty\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "f206b6c75a13f92bc97dfa1e930151b21df17184be9d9c72a2ae387f01f205b5"
  },
  "kernelspec": {
   "display_name": "Python 3.8.13 ('FYP')",
   "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": "3.8.13"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}