classification.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Importing libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import json\n",
    "import seaborn as sns\n",
    "import re\n",
    "import csv\n",
    "import matplotlib.pyplot as plt\n",
    "from tqdm import tqdm\n",
    "\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn import naive_bayes, svm\n",
    "from sklearn.model_selection import cross_validate\n",
    "from sklearn.metrics import classification_report, accuracy_score\n",
    "from sklearn.feature_selection import SelectKBest\n",
    "from sklearn.feature_selection import chi2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Extract data from file keeping only 4 columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "16559it [00:00, 23027.71it/s]\n",
      "100%|██████████| 16559/16559 [00:00<00:00, 1086636.83it/s]\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>book_id</th>\n",
       "      <th>book_name</th>\n",
       "      <th>genre</th>\n",
       "      <th>summary</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>620</td>\n",
       "      <td>Animal Farm</td>\n",
       "      <td>{\"/m/016lj8\": \"Roman \\u00e0 clef\", \"/m/06nbt\":...</td>\n",
       "      <td>Old Major, the old boar on the Manor Farm, ca...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>843</td>\n",
       "      <td>A Clockwork Orange</td>\n",
       "      <td>{\"/m/06n90\": \"Science Fiction\", \"/m/0l67h\": \"N...</td>\n",
       "      <td>Alex, a teenager living in near-future Englan...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  book_id           book_name  \\\n",
       "0     620         Animal Farm   \n",
       "1     843  A Clockwork Orange   \n",
       "\n",
       "                                               genre  \\\n",
       "0  {\"/m/016lj8\": \"Roman \\u00e0 clef\", \"/m/06nbt\":...   \n",
       "1  {\"/m/06n90\": \"Science Fiction\", \"/m/0l67h\": \"N...   \n",
       "\n",
       "                                             summary  \n",
       "0   Old Major, the old boar on the Manor Farm, ca...  \n",
       "1   Alex, a teenager living in near-future Englan...  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = []\n",
    "\n",
    "with open(\"Data/booksummaries/booksummaries.txt\", encoding=\"utf8\") as f:\n",
    "    reader = csv.reader(f, dialect='excel-tab')\n",
    "    for row in tqdm(reader):\n",
    "        data.append(row)\n",
    "\n",
    "book_id = []\n",
    "book_name = []\n",
    "summary = []\n",
    "genre = []\n",
    "\n",
    "for i in tqdm(data):\n",
    "    book_id.append(i[0])\n",
    "    book_name.append(i[2])\n",
    "    genre.append(i[5])\n",
    "    summary.append(i[6])\n",
    "\n",
    "books = pd.DataFrame({'book_id': book_id, 'book_name': book_name,\n",
    "                       'genre': genre, 'summary': summary})\n",
    "\n",
    "books.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dataset size before cleaning: 16559\n",
      "Dataset size after cleaning: 12841\n"
     ]
    }
   ],
   "source": [
    "# json genres example\n",
    "json.loads(books['genre'][1]).values()\n",
    "\n",
    "print(f'Dataset size before cleaning: {len(books)}')\n",
    "\n",
    "# drop data where genre is empty\n",
    "books.drop(books[books['genre']==''].index, inplace=True)\n",
    "\n",
    "# drop data where summary is empty\n",
    "books.drop(books[books['summary']==''].index, inplace=True)\n",
    "\n",
    "\n",
    "print(f'Dataset size after cleaning: {len(books)}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dataset size after cleaning: 11941\n"
     ]
    }
   ],
   "source": [
    "s = books.summary.str.len().sort_values().index\n",
    "books = books.reindex(s)[900:]\n",
    "print(f'Dataset size after cleaning: {len(books)}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Decode JSON format in 'genre' column to 'new_genre' column\\\n",
    "Create a dictionary with all genres as keys and their number of occurrence as values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "genres = []\n",
    "for i in books['genre']:\n",
    "    x = list(json.loads(i).values())\n",
    "    genres.append(x)\n",
    "books['genre_new'] = genres\n",
    "\n",
    "genres_sum = sum(genres,[])\n",
    "all_genres = {}\n",
    "\n",
    "for g in set(genres_sum):\n",
    "    all_genres[g] = genres_sum.count(g)\n",
    "\n",
    "all_genres = dict(sorted(all_genres.items(), key=lambda x: x[1]))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Different genres count: 226\n",
      "Top 40 popular genres:\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 864x1080 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(f'Different genres count: {len(all_genres)}')\n",
    "print('Top 40 popular genres:')\n",
    "vis = pd.DataFrame({'genre':list(all_genres.keys()), 'count':list(all_genres.values())}) \n",
    "\n",
    "# visualize genres and frequencies\n",
    "dat = vis.nlargest(columns=\"count\", n = 40) \n",
    "plt.figure(figsize=(12,15)) \n",
    "ax = sns.barplot(data=dat, x= \"count\", y = \"genre\") \n",
    "ax.set(ylabel = 'Genre') \n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Remove genres number: 221\n",
      "All genres number: 5\n",
      "{'Novel', 'Fantasy', 'Crime Fiction', 'Mystery', 'Thriller'}\n"
     ]
    }
   ],
   "source": [
    "# remove_genres = list({key: value for key, value in all_genres.items() if value <= 700})\n",
    "# remove_genres.append(\"Speculative fiction\")\n",
    "# all_genres = {key: value for key, value in all_genres.items() if value > 700}\n",
    "# all_genres.pop('Speculative fiction')\n",
    "genres_list = {\"Fantasy\", \"Novel\", \"Crime Fiction\", \"Thriller\", \"Mystery\"}\n",
    "remove_genres = all_genres\n",
    "for g in genres_list: \n",
    "    remove_genres.pop(g)\n",
    "all_genres = genres_list\n",
    "\n",
    "print(f'Remove genres number: {len(remove_genres)}')\n",
    "print(f'All genres number: {len(all_genres)}')\n",
    "print(all_genres)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5895\n"
     ]
    }
   ],
   "source": [
    "# remove genres from dataset\n",
    "for rem in remove_genres:\n",
    "    i = 0\n",
    "    for g in books['genre_new']:\n",
    "        i += 1\n",
    "        if rem in g: \n",
    "            g.remove(rem)\n",
    "\n",
    "id_empty=[]\n",
    "for id in books['book_id']:\n",
    "    if books[books['book_id']==id]['genre_new'].tolist()[0] == []: \n",
    "        id_empty.append(id)\n",
    "\n",
    "print(len(id_empty))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# drop data where genre is empty\n",
    "for id in id_empty:\n",
    "    books.drop(books[books['book_id']==id].index, inplace=True)\n",
    "\n",
    "genres = []\n",
    "for i in books['genre_new']:\n",
    "    x = list(i)[0]\n",
    "    genres.append(x)\n",
    "books['genre'] = genres"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Preprocessing text functions:\n",
    "- to lowercase\n",
    "- remove everething except letters\n",
    "- remove whitespaces \n",
    "- remove stop words\n",
    "- text stemming (Porter)\n",
    "<img src=\"https://user.oc-static.com/upload/2020/10/22/16033551309054_stemming%20example.png\" alt=\"drawing\" width=\"300\"/>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def preprocess(text):\n",
    "    text = re.sub(\"\\'\", \"\", str(text)) \n",
    "    text = re.sub(\"[^a-zA-Z]\", \" \", str(text)) \n",
    "    text = ' '.join(text.split()) \n",
    "    text = text.lower() \n",
    "\n",
    "    return text\n",
    "\n",
    "books['clear_summary'] = books['summary'].apply(lambda x: preprocess(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Removing stop words (\"a\", \"an\"...)\n",
    "from nltk.corpus import stopwords\n",
    "stop_words = set(stopwords.words('english'))\n",
    "def remove_stopwords(text):\n",
    "    no_stopword_text = [w for w in text.split() if not w in stop_words]\n",
    "    return ' '.join(no_stopword_text)\n",
    "\n",
    "books['clear_summary'] = books['clear_summary'].apply(lambda x: remove_stopwords(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "from nltk.stem import PorterStemmer\n",
    "stemmer = PorterStemmer()\n",
    "def stemming(text):\n",
    "    stemText = \"\"\n",
    "    for word in text.split():\n",
    "        stem = stemmer.stem(word)\n",
    "        stemText += stem\n",
    "        stemText += \" \"\n",
    "    stemText = stemText.strip()\n",
    "    return stemText\n",
    "\n",
    "books['clear_summary'] = books['clear_summary'].apply(lambda x: stemming(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "data size: 6046\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>book_id</th>\n",
       "      <th>book_name</th>\n",
       "      <th>summary</th>\n",
       "      <th>genre_new</th>\n",
       "      <th>clear_summary</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>genre</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Crime Fiction</th>\n",
       "      <td>606</td>\n",
       "      <td>606</td>\n",
       "      <td>606</td>\n",
       "      <td>606</td>\n",
       "      <td>606</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Fantasy</th>\n",
       "      <td>2144</td>\n",
       "      <td>2144</td>\n",
       "      <td>2144</td>\n",
       "      <td>2144</td>\n",
       "      <td>2144</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Mystery</th>\n",
       "      <td>750</td>\n",
       "      <td>750</td>\n",
       "      <td>750</td>\n",
       "      <td>750</td>\n",
       "      <td>750</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Novel</th>\n",
       "      <td>2004</td>\n",
       "      <td>2004</td>\n",
       "      <td>2004</td>\n",
       "      <td>2004</td>\n",
       "      <td>2004</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Thriller</th>\n",
       "      <td>542</td>\n",
       "      <td>542</td>\n",
       "      <td>542</td>\n",
       "      <td>542</td>\n",
       "      <td>542</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               book_id  book_name  summary  genre_new  clear_summary\n",
       "genre                                                               \n",
       "Crime Fiction      606        606      606        606            606\n",
       "Fantasy           2144       2144     2144       2144           2144\n",
       "Mystery            750        750      750        750            750\n",
       "Novel             2004       2004     2004       2004           2004\n",
       "Thriller           542        542      542        542            542"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# After summary preprocessing\n",
    "print(f'data size: {len(books)}')\n",
    "books.groupby('genre').count()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Encode target labels with value between 0 and n_classes-1."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['Crime Fiction', 'Fantasy', 'Mystery', 'Novel', 'Thriller'],\n",
       "      dtype=object)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.preprocessing import LabelEncoder\n",
    "\n",
    "encoder = LabelEncoder()\n",
    "y = encoder.fit_transform(books['genre'])\n",
    "encoder.inverse_transform([0,1,2,3,4])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Convert text to numbers using TFIDF (term frequency–inverse document frequency) approach\n",
    "\n",
    "+ **Term frequency** = (Number of Occurrences of a word)/(Total words in the document)\n",
    "+ **IDF(word)** = Log((Total number of documents)/(Number of documents containing the word))\n",
    "\n",
    "*max_features* - equal to 4000, which means we want to use the 4000 most frequently occurring words as features to train our classifier.\\\n",
    "*min_df* - ignore terms that have a document frequency strictly lower than the given number\\\n",
    "*max_df* - include only those words that appear in a maximum of 65% of all documents\\\n",
    "(Words that appear in almost every document are usually not suitable for classification because they do not provide any unique information about the document)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 142,
   "metadata": {},
   "outputs": [],
   "source": [
    "tfidfconverter = TfidfVectorizer(max_features=6000, min_df=10, max_df=0.65)\n",
    "X = tfidfconverter.fit_transform(books['clear_summary'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Chi-square feature selection "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 143,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(6046, 6000)\n",
      "(6046, 1500)\n"
     ]
    }
   ],
   "source": [
    "# K features with highest chi-squared statistics are selected\n",
    "chi2_features = SelectKBest(chi2, k = 1500)\n",
    "print(X.shape)\n",
    "X = chi2_features.fit_transform(X, y)\n",
    "print(X.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Splitting data into training and test sets (90%/10%)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, shuffle=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "1. Naive Bayes Classifier Algorithm & model evaluation\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.73211009 0.72242647 0.69669118 0.71691176 0.70772059 0.70772059\n",
      " 0.73161765 0.69669118 0.73345588 0.68566176]\n",
      "best model: 8\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.91      0.52      0.67        61\n",
      "           1       0.82      0.84      0.83       216\n",
      "           2       0.87      0.44      0.58        78\n",
      "           3       0.58      0.92      0.71       192\n",
      "           4       1.00      0.09      0.16        58\n",
      "\n",
      "    accuracy                           0.71       605\n",
      "   macro avg       0.84      0.56      0.59       605\n",
      "weighted avg       0.78      0.71      0.68       605\n",
      "\n",
      "Accurancy 0.7090909090909091\n",
      "[1 3 2 0 3 3 0 2 1 1 3 3 1 3 3]\n",
      "[1 3 1 0 3 3 1 2 1 1 3 3 1 3 3]\n"
     ]
    }
   ],
   "source": [
    "# fitting the classification model\n",
    "classifier_nb = naive_bayes.MultinomialNB()\n",
    "\n",
    "# x10 cross validation\n",
    "cv_results = cross_validate(classifier_nb, X_train, y_train, cv=10, return_estimator=True)\n",
    "print(cv_results['test_score'])\n",
    "index_best = np.argmax(cv_results['test_score'])\n",
    "print(f'best model: {index_best}')\n",
    "classifier_nb = cv_results['estimator'][index_best]\n",
    "\n",
    "# predict genre on test dataset\n",
    "y_pred = classifier_nb.predict(X_test)\n",
    "\n",
    "# Model evaluation\n",
    "#print(confusion_matrix(y_test,y_pred))\n",
    "print(classification_report(y_test,y_pred))\n",
    "print(f'Accurancy {accuracy_score(y_test, y_pred)}')\n",
    "\n",
    "print(y_test[:15])\n",
    "print(y_pred[:15])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Testing hiper-parametr alpha of the MultinomialNB classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best alpha:  {'alpha': 0.1}\n",
      "Best score:  0.7516945493793848\n",
      "Accurancy 0.7487603305785124\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import GridSearchCV\n",
    "c_nb = naive_bayes.MultinomialNB()\n",
    "\n",
    "# Grid search with x10 cross validation\n",
    "params = [{'alpha': [0.0001, 0.001, 0.01, 0.1, 0.5, 1]}]\n",
    "gs = GridSearchCV(c_nb,\n",
    "                param_grid=params,\n",
    "                cv=10)\n",
    "gs.fit(X_train, y_train)\n",
    "print(\"Best alpha: \", gs.best_params_)\n",
    "print(\"Best score: \", gs.best_score_)\n",
    "gs_y = gs.predict(X_test)\n",
    "print(f'Accurancy {accuracy_score(y_test, gs_y)}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "2. SVM algorithm using the RBF (Gaussian) kernel (Gamma = 1, default C = 10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.72451791 0.68595041 0.73002755 0.70247934 0.74104683 0.74655647\n",
      " 0.72727273 0.75414365 0.73480663 0.72375691]\n",
      "best model: 7\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.62      0.56      0.59       243\n",
      "           1       0.82      0.83      0.82       869\n",
      "           2       0.65      0.58      0.61       282\n",
      "           3       0.70      0.79      0.74       802\n",
      "           4       0.55      0.44      0.49       223\n",
      "\n",
      "    accuracy                           0.72      2419\n",
      "   macro avg       0.67      0.64      0.65      2419\n",
      "weighted avg       0.72      0.72      0.72      2419\n",
      "\n",
      "Accurancy 0.7209590739975197\n"
     ]
    }
   ],
   "source": [
    "# fitting the classification model\n",
    "classifier_svc = svm.SVC(C = 10, kernel='rbf',gamma=1)\n",
    "\n",
    "# x10 cross validation\n",
    "cv_results = cross_validate(classifier_svc, X_train, y_train, cv=10, return_estimator=True)\n",
    "print(cv_results['test_score'])\n",
    "index_best = np.argmax(cv_results['test_score'])\n",
    "print(f'best model: {index_best}')\n",
    "classifier_svc = cv_results['estimator'][index_best]\n",
    "\n",
    "y_pred=classifier_svc.predict(X_test)\n",
    "\n",
    "# Model evaluation\n",
    "#print(confusion_matrix(y_test,y_pred))\n",
    "print(classification_report(y_test,y_pred))\n",
    "print(f'Accurancy {accuracy_score(y_test, y_pred)}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "SVM hyperparameters: Gamma = 0.1, C = 10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.73211009 0.74264706 0.71875    0.73713235 0.73161765 0.73161765\n",
      " 0.73345588 0.72977941 0.73897059 0.75      ]\n",
      "best model: 9\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.72      0.65      0.68        66\n",
      "           1       0.84      0.90      0.87       228\n",
      "           2       0.70      0.46      0.55        81\n",
      "           3       0.71      0.78      0.74       187\n",
      "           4       0.59      0.60      0.60        43\n",
      "\n",
      "    accuracy                           0.75       605\n",
      "   macro avg       0.71      0.68      0.69       605\n",
      "weighted avg       0.75      0.75      0.75       605\n",
      "\n",
      "Accurancy 0.7537190082644628\n"
     ]
    }
   ],
   "source": [
    "# fitting the classification model\n",
    "classifier_svc = svm.SVC(C = 10, kernel='rbf',gamma=0.1)\n",
    "\n",
    "# x5 cross validation\n",
    "cv_results = cross_validate(classifier_svc, X_train, y_train, cv=10, return_estimator=True)\n",
    "print(cv_results['test_score'])\n",
    "index_best = np.argmax(cv_results['test_score'])\n",
    "print(f'best model: {index_best}')\n",
    "classifier_svc = cv_results['estimator'][index_best]\n",
    "\n",
    "y_pred=classifier_svc.predict(X_test)\n",
    "\n",
    "# Model evaluation\n",
    "#print(confusion_matrix(y_test,y_pred))\n",
    "print(classification_report(y_test,y_pred))\n",
    "print(f'Accurancy {accuracy_score(y_test, y_pred)}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "SVM hyperparameters: Gamma = 10, C = 10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.3559633  0.35294118 0.35110294 0.35110294 0.35294118 0.35294118\n",
      " 0.35477941 0.35477941 0.35477941 0.35477941]\n",
      "best: 0\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.00      0.00      0.00        66\n",
      "           1       0.38      1.00      0.55       228\n",
      "           2       0.00      0.00      0.00        81\n",
      "           3       1.00      0.01      0.01       187\n",
      "           4       0.00      0.00      0.00        43\n",
      "\n",
      "    accuracy                           0.38       605\n",
      "   macro avg       0.28      0.20      0.11       605\n",
      "weighted avg       0.45      0.38      0.21       605\n",
      "\n",
      "Accurancy 0.3785123966942149\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\Users\\Asus\\AppData\\Local\\Programs\\Python\\Python38-32\\lib\\site-packages\\sklearn\\metrics\\_classification.py:1327: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "c:\\Users\\Asus\\AppData\\Local\\Programs\\Python\\Python38-32\\lib\\site-packages\\sklearn\\metrics\\_classification.py:1327: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "c:\\Users\\Asus\\AppData\\Local\\Programs\\Python\\Python38-32\\lib\\site-packages\\sklearn\\metrics\\_classification.py:1327: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    }
   ],
   "source": [
    "# fitting the classification model\n",
    "classifier_svc = svm.SVC(C = 10, kernel='rbf',gamma=10)\n",
    "\n",
    "# x5 cross validation\n",
    "cv_results = cross_validate(classifier_svc, X_train, y_train, cv=10, return_estimator=True)\n",
    "print(cv_results['test_score'])\n",
    "index_best = np.argmax(cv_results['test_score'])\n",
    "print(f'best: {index_best}')\n",
    "classifier_svc = cv_results['estimator'][index_best]\n",
    "\n",
    "y_pred=classifier_svc.predict(X_test)\n",
    "\n",
    "# Model evaluation\n",
    "#print(confusion_matrix(y_test,y_pred))\n",
    "print(classification_report(y_test,y_pred))\n",
    "print(f'Accurancy {accuracy_score(y_test, y_pred)}')"
   ]
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "29c7a3964a9838c84f1d19dc4c2535ace62f5c920b4276f4747fa80ddec8638a"
  },
  "kernelspec": {
   "display_name": "Python 3.8.5 32-bit",
   "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": "3.8.5"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}