co-training-results.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Co-training\n",
    "\n",
    "**Autori:** Peter Macinec, Lukas Janik, Vajk Pomichal, Frantisek Sefcik"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Zakladne nastavenia a import kniznic"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "\n",
    "# plots\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "import json\n",
    "\n",
    "import re\n",
    "\n",
    "import nltk\n",
    "from nltk.stem.porter import PorterStemmer\n",
    "import string\n",
    "from nltk.corpus import stopwords\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn import model_selection as ms\n",
    "\n",
    "from sklearn.metrics import confusion_matrix\n",
    "\n",
    "import statistics\n",
    "\n",
    "from functools import reduce\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore', category=FutureWarning)\n",
    "warnings.filterwarnings('ignore', category=UserWarning)\n",
    "pd.options.mode.chained_assignment = None  # default='warn'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Nacitanie datasetu\n",
    "\n",
    "Nase data su dostupne v dvoch suboroch, *train.tsv* a *test.tsv*. Nacitame ich oba a vykoname na nich zakladnu analyzu. Zdroj: https://www.kaggle.com/c/stumbleupon"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# trenovacie data\n",
    "df = pd.read_csv('data/train.tsv', sep='\\t')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# testovacie data\n",
    "df_t = pd.read_csv('data/test.tsv', sep='\\t')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Textove atributy"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Najskor predspracujeme text. Ziskame ho z atributu boilerplate:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['body_content'] = df['boilerplate'].apply(lambda x: json.loads(x)['body'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Teraz odstranime vsetky znaky, ktore nie su znaky slov. Pouzijeme na to regularne vyrazy:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['body_content'].replace({r'[^\\x00-\\x7F]+':''}, regex=True, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['body_content'] = df['body_content'].apply(lambda x: re.sub('[^a-zA-Z]', ' ', str(x)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Este ako male upravy, aby nas slovnik obsahoval kazde slovo len raz, dame ich vsetky na lowercase a rozdelime texty na slova, aby sme ich nasledne mohli spracovat:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['body_content'] = df['body_content'].apply(lambda x: str(x).lower().split())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Teraz potrebujeme este odstranit slova, ktore nedavaju vyznam. O jednom raze prevedieme slova na ich korenovy zaklad pouzitim stemmingu:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "porter_stemmer = PorterStemmer()\n",
    "stopwords = set(stopwords.words('english'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['body_content'] = df['body_content'].apply(lambda x: [porter_stemmer.stem(word) for word in x if word not in stopwords])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Teraz uz mame vsetky slova pripravene, uz ich len naspat spojime do jednej suvislej vety, aby s nimi vedeli lahsie pracovat algoritmy spracovania textu:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['body_content_final'] = df['body_content'].apply(lambda x: ' '.join(x))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### TF-IDF\n",
    "\n",
    "Kedze mame ako atribut text stranky, TF-IDF nam moze pomoct. Vyberieme intuitivne top 1000 slov:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "tv = TfidfVectorizer(max_features = 1000)\n",
    "tf_idf = tv.fit_transform(df['body_content_final']).toarray()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Atribut URL\n",
    "\n",
    "Dalsi dobry ukazovatel moze byt URL. Tak isto, natrenujeme na nej TF-IDF s top 1000 slovami. Najskor ich ale musim predspracovat:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['url_new'] = df['url'].apply(lambda x: re.sub('[^a-zA-Z]', ' ', str(x)))\n",
    "df['url_new'] = df['url_new'].apply(lambda x: str(x).lower().split())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['url_new'] = df['url_new'].apply(lambda x: [porter_stemmer.stem(word) for word in x if word not in stopwords])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['url_final'] = df['url_new'].apply(lambda x: ' '.join(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "tv_url = TfidfVectorizer(max_features = 1000)\n",
    "tf_idf_url = tv_url.fit_transform(df['url_final']).toarray()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Numericke atributy\n",
    "\n",
    "Ako posledne atributy vyberieme vsetky numericke, ktore by nam mohli nieco prezradit. Pouzijeme ich spolu s TF-IDF crtami z URL."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_feature_set = ['avglinksize', 'commonlinkratio_1', 'commonlinkratio_2', 'commonlinkratio_3', 'commonlinkratio_4',\n",
    "                   'hasDomainLink','lengthyLinkDomain','linkwordscore','numberOfLinks',\n",
    "                   'numwords_in_url', 'parametrizedLinkRatio']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Benchmark model\n",
    "\n",
    "Natrenujeme benchmarkovy model, ktory bude natrenovany na vsetkych atributoch. Jeho vysledky sa nasledne budeme snazit dosiahnut s minimom oznacenych dat s co-trainingom."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "y = df.label"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Postupne vyberieme vsetky crty, ktore sme popisali vyssie. Aj TF-IDF z textov, aj z URL, aj numericke atributy:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "df1 = df.loc[:, num_feature_set]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "df2 = pd.DataFrame(tf_idf_url)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_temp = pd.concat([df2, pd.DataFrame(tf_idf)], axis=1, join_axes=[df2.index])\n",
    "X_temp.columns = list(range(0,2000))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = pd.concat([df1, X_temp], axis=1, join_axes=[df1.index])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Data uz mame spojene, podme sa na ne pozriet:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "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>avglinksize</th>\n",
       "      <th>commonlinkratio_1</th>\n",
       "      <th>commonlinkratio_2</th>\n",
       "      <th>commonlinkratio_3</th>\n",
       "      <th>commonlinkratio_4</th>\n",
       "      <th>hasDomainLink</th>\n",
       "      <th>lengthyLinkDomain</th>\n",
       "      <th>linkwordscore</th>\n",
       "      <th>numberOfLinks</th>\n",
       "      <th>numwords_in_url</th>\n",
       "      <th>...</th>\n",
       "      <th>1990</th>\n",
       "      <th>1991</th>\n",
       "      <th>1992</th>\n",
       "      <th>1993</th>\n",
       "      <th>1994</th>\n",
       "      <th>1995</th>\n",
       "      <th>1996</th>\n",
       "      <th>1997</th>\n",
       "      <th>1998</th>\n",
       "      <th>1999</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2.055556</td>\n",
       "      <td>0.676471</td>\n",
       "      <td>0.205882</td>\n",
       "      <td>0.047059</td>\n",
       "      <td>0.023529</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>170</td>\n",
       "      <td>8</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.157927</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.028198</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.064583</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3.677966</td>\n",
       "      <td>0.508021</td>\n",
       "      <td>0.288770</td>\n",
       "      <td>0.213904</td>\n",
       "      <td>0.144385</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>40</td>\n",
       "      <td>187</td>\n",
       "      <td>9</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.031715</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2.382883</td>\n",
       "      <td>0.562016</td>\n",
       "      <td>0.321705</td>\n",
       "      <td>0.120155</td>\n",
       "      <td>0.042636</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>55</td>\n",
       "      <td>258</td>\n",
       "      <td>11</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1.543103</td>\n",
       "      <td>0.400000</td>\n",
       "      <td>0.100000</td>\n",
       "      <td>0.016667</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>24</td>\n",
       "      <td>120</td>\n",
       "      <td>5</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2.676471</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>0.222222</td>\n",
       "      <td>0.123457</td>\n",
       "      <td>0.043210</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>14</td>\n",
       "      <td>162</td>\n",
       "      <td>10</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.02762</td>\n",
       "      <td>0.107720</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.129191</td>\n",
       "      <td>0.128223</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.088102</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 2011 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   avglinksize  commonlinkratio_1  commonlinkratio_2  commonlinkratio_3  \\\n",
       "0     2.055556           0.676471           0.205882           0.047059   \n",
       "1     3.677966           0.508021           0.288770           0.213904   \n",
       "2     2.382883           0.562016           0.321705           0.120155   \n",
       "3     1.543103           0.400000           0.100000           0.016667   \n",
       "4     2.676471           0.500000           0.222222           0.123457   \n",
       "\n",
       "   commonlinkratio_4  hasDomainLink  lengthyLinkDomain  linkwordscore  \\\n",
       "0           0.023529              0                  1             24   \n",
       "1           0.144385              0                  1             40   \n",
       "2           0.042636              0                  1             55   \n",
       "3           0.000000              0                  0             24   \n",
       "4           0.043210              0                  1             14   \n",
       "\n",
       "   numberOfLinks  numwords_in_url  ...   1990     1991      1992  1993  \\\n",
       "0            170                8  ...    0.0  0.00000  0.157927   0.0   \n",
       "1            187                9  ...    0.0  0.00000  0.031715   0.0   \n",
       "2            258               11  ...    0.0  0.00000  0.000000   0.0   \n",
       "3            120                5  ...    0.0  0.00000  0.000000   0.0   \n",
       "4            162               10  ...    0.0  0.02762  0.107720   0.0   \n",
       "\n",
       "       1994      1995  1996  1997      1998  1999  \n",
       "0  0.000000  0.028198   0.0   0.0  0.064583   0.0  \n",
       "1  0.000000  0.000000   0.0   0.0  0.000000   0.0  \n",
       "2  0.000000  0.000000   0.0   0.0  0.000000   0.0  \n",
       "3  0.000000  0.000000   0.0   0.0  0.000000   0.0  \n",
       "4  0.129191  0.128223   0.0   0.0  0.088102   0.0  \n",
       "\n",
       "[5 rows x 2011 columns]"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Vsetko vyzera byt v poriadku a pripravene. Este je potrebne pomenovat stlpce unikatnymi nazvami, co vyzaduju niektore algoritmy:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "col_names1 = list(range(0,1000))\n",
    "col_names1 = num_feature_set + col_names1\n",
    "col_names2 = list(range(1000,2000))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Rozdelime data na trenovaciu a testovaciu mnozinu v pomere 80% a 20%:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = ms.train_test_split(X, y, test_size=0.2, random_state=42)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ako benchmarkovy model sme vybrali RandomForest. Vyskusali sme aj dalsie klasifikatory, ale RandomForest sa ukazal ako najlepsi z nich. Parametre boli zvolene najskor intuitivne, potom vyladene."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "benchmark_clf = RandomForestClassifier(n_estimators=500, max_depth=20,\n",
    "                              random_state=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "benchmark_clf.fit(X_train, y_train)\n",
    "y_pred = benchmark_clf.predict(X_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Podme sa pozriet na vysledky predikcie:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[656,  74],\n",
       "       [221, 528]], dtype=int64)"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_matrix(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8005409060175794"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Accuracy 80% nie je zly vysledok. Teraz vieme, aky je strop a k comu sa chceme priblizit s co-trainingom a malym poctom oznacenych dat."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Co-training zhodnotenie\n",
    "\n",
    "Benchmark model uz mame natrenovany, tak este vyskusame co-training. Pouzijeme obe varianty, aj s iteraciami aj s thresholdom."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "class CombinedClassifier:\n",
    "    \n",
    "    def __init__(self,clf1, clf2, cols1, cols2):\n",
    "        self.clf1 = clf1\n",
    "        self.clf2 = clf2\n",
    "        self.cols1 = cols1\n",
    "        self.cols2 = cols2\n",
    "        \n",
    "    def predict(self, df):\n",
    "        result = []\n",
    "        predict1 = self.clf1.predict_proba(df[self.cols1])\n",
    "        predict2 = self.clf2.predict_proba(df[self.cols2])\n",
    "        \n",
    "        for i in range(len(df)):\n",
    "            prob0 = predict1[i][0] * predict2[i][0]\n",
    "            prob1 = predict1[i][1] * predict2[i][1]\n",
    "            res = 0 if prob0 > prob1 else 1\n",
    "            result.append(res)\n",
    "        return result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Co-training s iteraciami\n",
    "\n",
    "Ako prvy vyskusame co-training s iteraciami, ktory pouzili uz povodni autori (s drobnymi zmenami)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "def co_training(L, U, col_names1, col_names2, clf1, clf2,u=75, k=30,p=1,n=3, label='label', X_test=None, y_test=None, random_state=None):\n",
    "    '''\n",
    "    L - labeled data\n",
    "    U - unlabeled data\n",
    "    col_names1 - columns for clf1\n",
    "    col_names2 - columns for clf2\n",
    "    clf1 - classificator1\n",
    "    clf2 - classificator2\n",
    "    u - number of unlabeled data for training\n",
    "    k - number of iteration\n",
    "    p - number of positive examles to add label per iteration\n",
    "    n - number of negative examles to add label per iteration\n",
    "    label - name of column in L whit label\n",
    "    '''\n",
    "    \n",
    "    k0 = k\n",
    "    pred_1 = []\n",
    "    pred_2 = []\n",
    "    while k > 0:\n",
    "        # step 1: Use L to train a classifier h1 that considers only the x1 portion of x\n",
    "        clf1 = clf1.fit(L[col_names1],  L[label])\n",
    "        \n",
    "        # step 2: Use L, to train a classifier h2 that considers only the x2 portion of x\n",
    "        clf2 = clf2.fit(L[col_names2], L[label])\n",
    "        \n",
    "        # step 3: Allow hl to label p positive and n negative examples from U\n",
    "        predicted_prob1 = clf1.predict_proba(U[col_names1])\n",
    "        top_positive1 = predicted_prob1[:,1].argsort()[-p:]\n",
    "        top_negative1 = predicted_prob1[:,0].argsort()[-n:]\n",
    "        \n",
    "        # step 4: Allow hl to label p positive and n negative examples from U\n",
    "        predicted_prob2 = clf2.predict_proba(U[col_names2])\n",
    "        top_positive2 = predicted_prob2[:,1].argsort()[-p:]\n",
    "        top_negative2 = predicted_prob2[:,0].argsort()[-n:]    \n",
    "        \n",
    "        # step 5: Add these self-labeled examples to L\n",
    "        positive_ind = U.iloc[np.unique(np.concatenate((top_positive1,top_positive2))),:].index\n",
    "        negative_ind = U.iloc[np.unique(np.concatenate((top_negative1,top_negative2))),:].index\n",
    "        self_labeled = U.loc[np.unique(np.concatenate((positive_ind,negative_ind))),:]\n",
    "        U.drop(self_labeled.index, inplace=True)\n",
    "        self_labeled.loc[positive_ind, label] = 1\n",
    "        self_labeled.loc[negative_ind, label] = 0\n",
    "        L = pd.concat([L, self_labeled])\n",
    "    \n",
    "        if (X_test is not None and y_test is not None):\n",
    "            pred = clf1.predict(X_test[col_names1])\n",
    "            pred_1.append(accuracy_score(y_test,pred))\n",
    "            \n",
    "            pred = clf2.predict(X_test[col_names2])\n",
    "            pred_2.append(accuracy_score(y_test,pred))\n",
    "\n",
    "        k -= 1\n",
    "    \n",
    "    return [clf1, clf2, pred_1, pred_2]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "rozdelime stlpce pre dva klasifikatory"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "col_names1 = list(range(0,1000))\n",
    "col_names1 = num_feature_set + col_names1\n",
    "col_names2 = list(range(1000,2000))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ako dva vstupne klasifikatory pouzijeme logisticke regresie. Dosahovali podobne vysledky ako ine klasifikatory, avsak su velmi efektivne. Pre otestovanie viacerych behov je to idealna volba.\n",
    "\n",
    "Pouzijeme na vstupe 20 (10 positive, 10 negative) olabovanych zaznamov:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iteration: 0\n",
      "iteration: 1\n",
      "iteration: 2\n",
      "iteration: 3\n",
      "iteration: 4\n",
      "iteration: 5\n",
      "iteration: 6\n",
      "iteration: 7\n",
      "iteration: 8\n",
      "iteration: 9\n",
      "iteration: 10\n",
      "iteration: 11\n",
      "iteration: 12\n",
      "iteration: 13\n",
      "iteration: 14\n",
      "iteration: 15\n",
      "iteration: 16\n",
      "iteration: 17\n",
      "iteration: 18\n",
      "iteration: 19\n",
      "iteration: 20\n",
      "iteration: 21\n",
      "iteration: 22\n",
      "iteration: 23\n",
      "iteration: 24\n",
      "iteration: 25\n",
      "iteration: 26\n",
      "iteration: 27\n",
      "iteration: 28\n",
      "iteration: 29\n"
     ]
    }
   ],
   "source": [
    "scores = [[0 for _ in range(3)] for _ in range(30)]\n",
    "\n",
    "for x in range(30):\n",
    "    df_x0 = X_train[y_train == 0].sample(10)\n",
    "    df_x1 = X_train[y_train == 1].sample(10)\n",
    "    \n",
    "    train = pd.concat([df_x0,df_x1])\n",
    "    x_train = X_train.drop(train.index)\n",
    "    train['label'] = y_train.loc[train.index]\n",
    "\n",
    "    clf1 = LogisticRegression(n_jobs=-1)\n",
    "    clf2 = LogisticRegression(n_jobs=-1)\n",
    "\n",
    "    new_1, new_2, scores[x][0], scores[x][1] = co_training(L=train,U=x_train, \n",
    "                               col_names1=col_names1,\n",
    "                               col_names2=col_names2, clf1=clf1, clf2=clf2, k=30, \n",
    "                                                           n=2, p=2, X_test=X_test, y_test=y_test)\n",
    "    \n",
    "    \n",
    "    print(\"iteration:\", x)\n",
    "\n",
    "    cl = CombinedClassifier(new_1,new_2,col_names1,col_names2)\n",
    "    pred = cl.predict(X_test)\n",
    "    scores[x][2] = accuracy_score(y_test, pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Nasledne zistime ako sa darilo klasifikatorom v uceni pocas co-trainingu:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "avg_accuracy1 = []\n",
    "avg_accuracy2 = []\n",
    "model_accuracy1 = []\n",
    "model_accuracy2 = []\n",
    "for i in range(30):\n",
    "    sum_1 = 0 \n",
    "    sum_2 = 0\n",
    "    for j in range(30):\n",
    "        sum_1 += scores[j][0][i]\n",
    "        sum_2 += scores[j][1][i]\n",
    "        if i == 29:\n",
    "            model_accuracy1.append(scores[j][0][i])\n",
    "            model_accuracy2.append(scores[j][1][i])\n",
    "    avg_accuracy1.append(sum_1 / 30)\n",
    "    avg_accuracy2.append(sum_2 / 30)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x20d9f0c9518>]"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(avg_accuracy1)\n",
    "plt.plot(avg_accuracy2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Jednotlive spustenia sa od seba znacne odlisuju, avsak stale mozeme vidiet trend a sice, lepsi klasifikator pomoze tomu horsiemu dosiahnut zlepsenie, avsak sam seba o nieco zhorsi. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pozrieme sa aka je accuracy 1. klasifikatora po 30 behoch:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: 0.4665314401622718\n",
      "Max: 0.6308316430020284\n",
      "Avg: 0.5946810908271355\n",
      "Std: 0.03790474472101761\n"
     ]
    }
   ],
   "source": [
    "print(\"Min: \" + str(np.min(model_accuracy1)))\n",
    "print(\"Max: \" + str(np.max(model_accuracy1)))\n",
    "print(\"Avg: \" + str(np.mean(model_accuracy1)))\n",
    "print(\"Std: \" + str(np.std(model_accuracy1)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pozrieme sa aka je accuracy 2. klasifikatora po 30 behoch:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: 0.6274509803921569\n",
      "Max: 0.7748478701825557\n",
      "Avg: 0.702839756592292\n",
      "Std: 0.034876359128010584\n"
     ]
    }
   ],
   "source": [
    "print(\"Min: \" + str(np.min(model_accuracy2)))\n",
    "print(\"Max: \" + str(np.max(model_accuracy2)))\n",
    "print(\"Avg: \" + str(np.mean(model_accuracy2)))\n",
    "print(\"Std: \" + str(np.std(model_accuracy2)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pozrieme sa na accuracy kombinovaneho klasifikatora:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: 0.5409060175794456\n",
      "Max: 0.6997971602434077\n",
      "Avg: 0.6435429344151454\n",
      "Std: 0.036474576622485315\n"
     ]
    }
   ],
   "source": [
    "score_combined = []\n",
    "\n",
    "for i in range(30):\n",
    "    score_combined.append(scores[i][2])\n",
    "\n",
    "print(\"Min: \" + str(np.min(score_combined)))\n",
    "print(\"Max: \" + str(np.max(score_combined)))\n",
    "print(\"Avg: \" + str(np.mean(score_combined)))\n",
    "print(\"Std: \" + str(np.std(score_combined)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "V priemere sa uspesnost pohybuje okolo 65%, ale vidime, ze su pomerne velke vykyvy medzi min a max. Pravdepodobne je to sposobena zavislostou na kvalite vstupnych dat, ktore su oznacene"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Vacsia vstupna mnozina oznacenych dat a vyssi pocet iteracii"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Skusme drobnu zmenu parametrov: \n",
    "- zvysime pocet vzoriek, ktore budeme vyberat pre pozitivnu a negativnu triedu (mohlo by to priniest stabilnejsie vysledky)\n",
    "- zvysime pocet iteracii"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(5916, 2011)"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iteration:  0\n",
      "iteration:  1\n",
      "iteration:  2\n",
      "iteration:  3\n",
      "iteration:  4\n",
      "iteration:  5\n",
      "iteration:  6\n",
      "iteration:  7\n",
      "iteration:  8\n",
      "iteration:  9\n",
      "iteration:  10\n",
      "iteration:  11\n",
      "iteration:  12\n",
      "iteration:  13\n",
      "iteration:  14\n",
      "iteration:  15\n",
      "iteration:  16\n",
      "iteration:  17\n",
      "iteration:  18\n",
      "iteration:  19\n",
      "iteration:  20\n",
      "iteration:  21\n",
      "iteration:  22\n",
      "iteration:  23\n",
      "iteration:  24\n",
      "iteration:  25\n",
      "iteration:  26\n",
      "iteration:  27\n",
      "iteration:  28\n",
      "iteration:  29\n"
     ]
    }
   ],
   "source": [
    "scores = [[0 for _ in range(3)] for _ in range(30)]\n",
    "\n",
    "for x in range(30):\n",
    "    df_x0 = X_train[y_train == 0].sample(30)\n",
    "    df_x1 = X_train[y_train == 1].sample(30)\n",
    "    \n",
    "    train = pd.concat([df_x0,df_x1])\n",
    "    x_train = X_train.drop(train.index)\n",
    "    train['label'] = y_train.loc[train.index]\n",
    "\n",
    "    clf1 = LogisticRegression(n_jobs=-1)\n",
    "    clf2 = LogisticRegression(n_jobs=-1)\n",
    "\n",
    "    new_1, new_2, scores[x][0], scores[x][1] = co_training(L=train,U=x_train, \n",
    "                               col_names1=col_names1,\n",
    "                               col_names2=col_names2, clf1=clf1, clf2=clf2, k=60, \n",
    "                                                           n=2, p=2, X_test=X_test, y_test=y_test)\n",
    "    \n",
    "    \n",
    "    print(\"iteration: \",x)\n",
    "\n",
    "    cl = CombinedClassifier(new_1,new_2,col_names1,col_names2)\n",
    "    pred = cl.predict(X_test)\n",
    "    scores[x][2] = accuracy_score(y_test, pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "avg_accuracy1 = []\n",
    "avg_accuracy2 = []\n",
    "model_accuracy1 = []\n",
    "model_accuracy2 = []\n",
    "for i in range(60):\n",
    "    sum_1 = 0 \n",
    "    sum_2 = 0\n",
    "    for j in range(30):\n",
    "        sum_1 += scores[j][0][i]\n",
    "        sum_2 += scores[j][1][i]\n",
    "        if i == 29:\n",
    "            model_accuracy1.append(scores[j][0][i])\n",
    "            model_accuracy2.append(scores[j][1][i])\n",
    "    avg_accuracy1.append(sum_1 / 30)\n",
    "    avg_accuracy2.append(sum_2 / 30)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x20d9f0f3dd8>]"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYQAAAD8CAYAAAB3u9PLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzt3XmYXFd95vHvr7be98VaWkvLbssb4KWRTWyDDdgIkgETHGITghkWTRLgIWHCYM/wJIwTnnEymQAz8TDjgDEQsAkO2BrACAfbiWO8qGXLWJIta7XVai2tVq/qtbp+88e5LZVaLXWp1VKpu9/P89ynqu49VXWO3bpv3XPOvdfcHRERkVi+KyAiImcHBYKIiAAKBBERiSgQREQEUCCIiEhEgSAiIoACQUREIgoEEREBFAgiIhJJ5LsCJ6O2ttaXLl2a72qIiMwo69atO+DudZOVm1GBsHTpUlpaWvJdDRGRGcXMXsulnLqMREQEUCCIiEhEgSAiIoACQUREIgoEEREBFAgiIhJRIIiICDBXA8EddjwJ6+8Pz0VEZGadmHbK3GHLL+Bf/wZanwvrtj4K7/07SBXnt24iInk2NwIhMwqbHoYn/xb2vQQVi+E9fwNDPfDLv4CObXDr/VC+IN81FRHJm7kRCN+7GbY9BjXnwfv+N7zxgxBPhm31F8E/fQLuuR5u+T40XJHfuoqI5MncGEO4/Da4+Vvwqefgst87EgYAy98NH38UEin41rth/ffDEYWIyBxjPoMGVZubm/20Xdzu0AH4x4/Aa09BcW0Iigt+C5a9DZJFp+c7RUTOADNb5+7Nk5WbG11GuSiphY88DC+vhld+Chsfghe+C8liOPft0HQDLLseqpbku6YiIqeFAiFbPAmXfCAs6SHY+SS88jPY/Ai88pNQpvpcOPf6EA4NzVB6Dpjlt94iItNAXUa5cIf2zbD9cdj2OOz8Nxg5FLYV18K8S+CcS2DeG2D+pVDbBLH4ma+niMgE1GU0ncyg/oKwXPWHkB6G3S2w59ew96UwlfW5v4fRoVA+WQLz3wQLLgtLUSX07oW+vdC7D/r2gWeg9nyouwDqlofnuZwL0bcfdj0HFQ1wzsVHD5CLiJwCBcJUJFKw5DfCMmY0DR1boG09tL0QlpZvQnrw6PcWVoZuJoBXfw6ZdLTBoHJxONI456Kwsz/nEiibD7ueCUcm25+AfRuy6lEUgqehGRZeAcuug+Lq09duEZnV1GV0Oo2OQPsrMDIQQqD0HEgWHr394PZQpn0z7N8E+zaFYPHM0Z8VT8GiK8P4xZKrobsVdj8fHam8GIInlggD4JfcDBe8BwrKzmx7ReSslGuXkQLhbDQyCAc2w76N0LUr/Ppf8hZIlUxcfnQkdF+9/DBs+BF074JEIZz/Lqi/GIb7YKQfhg+F56PpMMYRS0RLPITVlf8hdEWJyKwyrYFgZiuBrwFx4Bvufte47V8Bro9eFgP17l4ZbRsFXoq2ve7u743WNwIPANXA88Dvu/vwieoxZwLhVGQy4TpNLz0IG38M/QdCOKRKwthGqgTiiVAukw6Lj0L37jBW0vwxuOZzUHZOvlsiItNk2gLBzOLAq8ANQCuwFrjV3Tcdp/xngMvc/WPR6z53L52g3D8CP3L3B8zs/wAvuvvXT1QXBcJJymQAz23GU9fr8C9/Hc7UThTAilVw9WdDSPS1w6H9YTC8/2AonygI3VjxZHgcGYDB7iPLUA8UVYUjlPoLw6B5InVamysiE5vOQHgL8CV3f1f0+g4Ad/9vxyn/K+DP3f3R6PUxgWBmBrQD89w9Pf47jkeBcAZ0bIMn7oKXfghMsTsxnoKCchjsOjJoHkuEa0nVnAflC6FiYfTYEAbTy+brfA6R02Q6p50uBHZlvW4FrjzOly4BGoHHslYXmlkLkAbucveHgBqgy93Hpti0Rt8z0WeuAlYBLF68OIfqyimpORc+8Pdw7edCl1NBeTQgXh+W4lrAYXQ4nLw3OhyWZDEUVoTyYwPn6WHo2BoGy/e/HB47tsL2f4Hh3qO/t7AymmF1cVjqlocjjIKy8JmpEgWGyGmWSyBM9K/weD8dbwEedPfsq8Mtdvc2M1sGPGZmLwE9uX6mu98D3APhCCGH+sp0qL8wLKcikYqm0F507LbBbuhpC2MXB7fD/o1hEP2Ffzhy0l82i4XAOeeSaJptc3gsm3dqdRSRw3IJhFZgUdbrBqDtOGVvAT6VvcLd26LH7Wb2BHAZ8E9ApZkloqOEE32mzEaFFWEZHzqZDHTthI7toctpqDeMRwz1wqH2cJ7Hr/7Xka6o8gZovDZMt112PZTWnfGmiMwWuQTCWqApmhW0m7DT/9D4Qma2HKgCns5aVwX0u/uQmdUCVwN/7e5uZo8DNxNmGt0GPHyqjZFZIBaD6mVhOZ6RgXCGeGtLmFH16hp48f6wbd4b4bx3hG6u4b5oqm005ba4OpzDsfiqcPa4iBwl12mn7wG+Sph2eq+7f9nM7gRa3H11VOZLQKG73571vt8A/i+QIdx74avu/s1o2zKOTDt9Afiwuw+dqB4aVJYJZUbDyXnbfhnO6N717NGD2anSMAZxqD2Md2Dh+lNLrgndTtXLoGppGLPQOIXMQjoxTeau4UPRQHfJ0VNdRwbCUcVrT4Vl11pIDxzZXlARLm9ec264SOHCK2DBpTrjW2Y8XdxO5q5UCTDBWd3JojDe0HhteJ0eggNboOs16Nx5ZNn9fJhhBYCFGU/z3nj0ZUfGtpXWh6mzFQ1QsShMpS045rQbkRlBgSBzV6IgdB3Nu+TYbYc6wgUKd6+Dtufh9WcgM3J0mcwo9HeEM72zFZRDSV00XTd6rDkPGt4cLpGuK9TKWUqBIDKRkhpoemdYTmQ0Db17wsUGu1vDdaT69oXLlPftD+dfbHsChrpD+UQRLLw8hEPl4jCTaqAzLP2dYSD88DWmoutMJQqjrqzoxL6ac8N4R65G0+Hih+mh6HEwhFLpOSEURSIKBJFTEU9A5aKwnEh3a7iPRevaMOj99N8dGfiOF4QZUEXVobvLx64zNRoeRw7BhgePvgJuUXX4zvKG6KzvBVC2IJzfkd391blz4vM6Dn9O1ZEr8VYtCbOwll6jixzOURpUFsmHkYFwVFBUFcY2JpMeDjv3jq1hObgtOiLZHU7wGzsCgXAUUrU0WpaE+4XHC8KRRiJ6HB2Kbta0N7p5074wnjLYFT6jamkIhoVXgMVDd9loOgzWZ9LhREGLhSMYi4UyZtHr+JFtmdHwXaMj0ZntI2EsZmzMpaIBSurDdGM5bTSoLHI2SxblFgRjEimoOz8sExnqhZ494WS/0vqpTZ/NZMIZ4zv/LSwv/yScOX66xVPhjPOS+jD2UlIbPdZFR05V0VIdZnyN9B85WXGwJ3SzxZPh8iljS6o43Pr2qDAaCgHlDviRx0ThketrzfEZZQoEkdmgoAzqTnFnFouFQe95bwi3is1koKc1+rWfjK5smwxjG+5hMN0z0U42M+75aHgei0dHJ6no6rgFYQeePebS3RqOcg61h+dtL4Tn4wfrz4SC8hAOlYvDJVfqo2tr1TbNickACgQRmVgsFnaM062oMiwTze4ak8lkDbh3wcDB8HywO4yzFJRHFz6Mlkw6nJE+0n/kZlAWCyGUHUaxRHR1NouOoiyU72k7Ekw9u+HgDtj22JGZZbFkGNAvrTv6iKW4OnR7jXXRZQ/2u4c2jHXrxeJRN9nC49/sKs8UCCJy9onFws42n/cIH7ta776N4V7mB14N04z3v3xkZtjhe6JHCitDiI4MhBA43oB+URQkNefBohVhmffGvB+FKBBERCZy1NV6f+fY7e7hiKV717iZXa+FMYymG46+90cmfXQ3WdeuMPNs44+i7yuEBZeHI6dkcTQBIGsywMW/fdoDUoEgIjIVZlndX2+Y+ud07w4XadwVLS/+IJwrMjru0m6N1ykQRERmtYqFUPF+uPj9R6/PZKIbUQ2Gx6LT332mQBARORvFYhArnOAaWqfxK8/YN4mIyFlNgSAiIoACQUREIgoEEREBFAgiIhJRIIiICKBAEBGRSE6BYGYrzWyzmW01s9sn2P4VM1sfLa+aWVe0/lIze9rMNprZr83sd7Pec5+Z7ch636XT1ywRETlZk56YZmZx4G7gBqAVWGtmq91901gZd/+TrPKfAS6LXvYDH3H3LWa2AFhnZmvcPboLB5939wenqS0iInIKcjlCWAFsdfft7j4MPAC87wTlbwXuB3D3V919S/S8DdgP1J1alUVE5HTIJRAWAruyXrdG645hZkuARuCxCbatAFLAtqzVX466kr5iZrrbt4hIHuUSCBPdi+94N2K+BXjQ/ehbHZnZfOC7wL93P3yn8DuAC4A3A9XAFyb8crNVZtZiZi3t7e05VFdERKYil0BoBRZlvW4A2o5T9hai7qIxZlYO/BT4ors/M7be3fd4MAR8i9A1dQx3v8fdm929ua5OvU0iIqdLLoGwFmgys0YzSxF2+qvHFzKz5UAV8HTWuhTwY+A77v7DceXnR48G3ARsmGojRETk1E06y8jd02b2aWANEAfudfeNZnYn0OLuY+FwK/CAu2d3J30QeCtQY2YfjdZ91N3XA98zszpCl9R64A+mpUUiIjIldvT+++zW3NzsLS0t+a6GiMiMYmbr3L15snI6U1lERAAFgoiIRBQIIiICKBBERCSiQBAREUCBICIiEQWCiIgACgQREYkoEEREBFAgiIhIRIEgIiKAAkFERCIKBBERARQIIiISUSCIiAigQBARkYgCQUREAAWCiIhEFAgiIgIoEEREJJJTIJjZSjPbbGZbzez2CbZ/xczWR8urZtaVte02M9sSLbdlrb/CzF6KPvN/mplNT5NERGQqEpMVMLM4cDdwA9AKrDWz1e6+aayMu/9JVvnPAJdFz6uBPweaAQfWRe/tBL4OrAKeAX4GrAQemaZ2iYjIScrlCGEFsNXdt7v7MPAA8L4TlL8VuD96/i7gUXc/GIXAo8BKM5sPlLv70+7uwHeAm6bcChEROWW5BMJCYFfW69Zo3THMbAnQCDw2yXsXRs8n/UwRETkzcgmEifr2/ThlbwEedPfRSd6b82ea2SozazGzlvb29kkrKyIiU5NLILQCi7JeNwBtxyl7C0e6i0703tbo+aSf6e73uHuzuzfX1dXlUF0REZmKXAJhLdBkZo1mliLs9FePL2Rmy4Eq4Oms1WuAG82sysyqgBuBNe6+B+g1s6ui2UUfAR4+xbaIiMgpmHSWkbunzezThJ17HLjX3Tea2Z1Ai7uPhcOtwAPRIPHYew+a2V8QQgXgTnc/GD3/Q+A+oIgwu0gzjERE8siy9t9nvebmZm9pacl3NUREZhQzW+fuzZOV05nKIiICKBBERCSiQBAREUCBICIiEQWCiIgACgQREYkoEEREBFAgiIhIRIEgIiKAAkFERCIKBBERARQIIiISUSCIiAigQBARkYgCQUREAAWCiIhEFAgiIgIoEEREJKJAEBERQIEgIiKRnALBzFaa2WYz22pmtx+nzAfNbJOZbTSz70frrjez9VnLoJndFG27z8x2ZG27dPqaJSIiJysxWQEziwN3AzcArcBaM1vt7puyyjQBdwBXu3unmdUDuPvjwKVRmWpgK/CLrI//vLs/OF2NERGRqcvlCGEFsNXdt7v7MPAA8L5xZT4J3O3unQDuvn+Cz7kZeMTd+0+lwiIicnrkEggLgV1Zr1ujddnOB843s6fM7BkzWznB59wC3D9u3ZfN7Ndm9hUzK8i51iIiMu1yCQSbYJ2Pe50AmoDrgFuBb5hZ5eEPMJsPvAFYk/WeO4ALgDcD1cAXJvxys1Vm1mJmLe3t7TlUV0REpiKXQGgFFmW9bgDaJijzsLuPuPsOYDMhIMZ8EPixu4+MrXD3PR4MAd8idE0dw93vcfdmd2+uq6vLoboiIjIVuQTCWqDJzBrNLEXo+lk9rsxDwPUAZlZL6ELanrX9VsZ1F0VHDZiZATcBG6bSABERmR6TzjJy97SZfZrQ3RMH7nX3jWZ2J9Di7qujbTea2SZglDB7qAPAzJYSjjD+ZdxHf8/M6ghdUuuBP5ieJomIyFSY+/jhgLNXc3Ozt7S05LsaIiIzipmtc/fmycrpTGUREQEUCCIiElEgiIgIoEAQEZGIAkFERAAFgoiIRBQIIiICKBBERCSiQBAREUCBICIiEQWCiIgACgQREYkoEEREBFAgiIhIRIEgIiKAAkFERCIKBBERARQIIiISUSCIiAigQBARkYgCQUREgBwDwcxWmtlmM9tqZrcfp8wHzWyTmW00s+9nrR81s/XRsjprfaOZPWtmW8zsB2aWOvXmiIjIVE0aCGYWB+4G3g1cBNxqZheNK9ME3AFc7e4XA3+ctXnA3S+Nlvdmrf8r4Cvu3gR0Ah8/taaIiMipyOUIYQWw1d23u/sw8ADwvnFlPgnc7e6dAO6+/0QfaGYGvB14MFr1beCmk6m4iIhMr1wCYSGwK+t1a7Qu2/nA+Wb2lJk9Y2Yrs7YVmllLtH5sp18DdLl7+gSfCYCZrYre39Le3p5DdUVEZCoSOZSxCdb5BJ/TBFwHNABPmtkl7t4FLHb3NjNbBjxmZi8BPTl8Zljpfg9wD0Bzc/OEZURE5NTlcoTQCizKet0AtE1Q5mF3H3H3HcBmQkDg7m3R43bgCeAy4ABQaWaJE3ymiIicQbkEwlqgKZoVlAJuAVaPK/MQcD2AmdUSupC2m1mVmRVkrb8a2OTuDjwO3By9/zbg4VNtjIiITN2kgRD1838aWAO8DPyju280szvNbGzW0Bqgw8w2EXb0n3f3DuBCoMXMXozW3+Xum6L3fAH4nJltJYwpfHM6GyYiIifHwo/1maG5udlbWlryXQ0RkRnFzNa5e/Nk5XSmsoiIAAoEERGJKBBERARQIIiISESBICIigAJBREQiCgQREQEUCCIiElEgiIgIoEAQEZGIAkFERIDc7ocgIiJT0NrZz3M7DvLcjoMcPDRMIm7EYzESMSMeM4pTcWpKCqgpTVFbmqKmtIDKoiTJeIxE3MJjzEgmYpSmEsRiE92eZvooEEREpqBncIS2rgF6B9McGkozMDzKoeFRegdHeGl3N89uP8jurgEAKoqSzK8oZDTjjGacdPR4aDhNV/9ITt/3z597K+fVl53OJikQRGRu6h9Os6d7kIHhUcoLk5QXJSgtSJCIh5703sERdncNsLtzgNbOAXZ3DbDrYD+7OvvZdXCA7oHj78irS1KsWFrNJ65t5MrGGi6YV3bcX/cjoxk6Dw3T3jdER98wXQMjpEczpEedkUz0OJqhrrTwtPx3yKZAEJEZaTTj9A2m6RkcYWBklMGRUQZHMtHjKL2DaboHRg4vPQMjtPcNsa9nkD3dg/QOpif83JJUnFjMjtmeSsRoqCpiUVUxly6qZFFVMQuriqgoSlKcSlCcilOSSlBcEKemJIVZbt07yXiM+vJC6stP/w5/MgoEETnjMhmnvW+Itq4B9nQPcmgoTUVRkqqSFJVFSSqKkyRjMXZ2HGJb+yG2t/exvf0Qrx3sp7t/mJ7BNH1DE+/QJ1JWmKCiKElNSYrG2hLesqyGeRVFzKsooDiVoHcwTe/gCD0DIWDSoxnmVxbRUFXEwsoiFlYVUVtScNr78PNNgSAiUzaacYbSowyNZBhKZ8LzdIbhdPilfqBvmL3dYaff1j3I3u4B2roG2dczSDqT+825EjFjcU0xjTUlXLygnLLCRNTNk6SsIEFRKk5hMk5RMk5hMkZhMk5pQQiB8qIk8Vm+I58uCgQROa6h9Chb9vXx8p4eXt7Ty8t7etja3kf/UJqhdCbnnXoqEWN+RSHzKwpZ0VgdnlcWMb+8kPmVhZQWJOgZSNPZH/rQu/uHGUpnWFJTwrK6EhZXF5OMa5b86aZAEJnj3J1t7YfYsLub1mjAdFdnP62dA7R1DRze6RcmYyyfV871y+soL0xSkIxRkIhTkIiRSoRf5al47PD6VCJGTUmKeRWFJ9WnLvmjQBCZY0ZGM2zd38dzOw7y7I4OnttxkAN9w4e315UV0FBVxKWLKnnvmxZwwfwyLpxfztKaEnW9zHI5BYKZrQS+BsSBb7j7XROU+SDwJcCBF939Q2Z2KfB1oBwYBb7s7j+Iyt8HvA3ojj7io+6+/pRaIzKLDaczPP96J/+25QDP7TxI72D6qP774fQoyXj4pV6cilOUCn3q6YzTMzBCz+AIvYNp+odHD3/mgopCrm2q48rGai5fUsXi6mIKk/E8tlLyadJAMLM4cDdwA9AKrDWz1e6+KatME3AHcLW7d5pZfbSpH/iIu28xswXAOjNb4+5d0fbPu/uD09kgkZnM3Tk0PEpH3xAH+oY50DfEroP9/GpbB89s76B/eJR4zHjDwgoaqoooSETdNskYqXiMkdEMAyOjDAyPMjAySv/wKIXJGPVlpUcNxC6oLOLKxmoWVRfnu8lyFsnlCGEFsNXdtwOY2QPA+4BNWWU+Cdzt7p0A7r4/enx1rIC7t5nZfqAO6EJkjnJ3eofSvN7Rz7b2Prbu7zv8+PrBfgZHMse8p7G2hA9c3sA1TbW85dwayguTeai5zHa5BMJCYFfW61bgynFlzgcws6cI3UpfcvefZxcwsxVACtiWtfrLZvZnwC+B2919aPyXm9kqYBXA4sWLc6iuyNnB3dm8r5fHX2lny75e9vYMsrd7kL09g0d128QMltSUcG5dKW9tqqOurIDa0rHr2xRQX15AfVn+T1qS2S+XQJhoFGn8XLME0ARcBzQAT5rZJWNdQ2Y2H/gucJu7j/38uQPYSwiJe4AvAHce80Xu90TbaW5uzn3iskge9A6O8Oz2gzy2eT9PvLKftu5BIPTVz68s4sL55Vx/QT3zygtZWFXEefWlLKkppiChfnvJv1wCoRVYlPW6AWiboMwz7j4C7DCzzYSAWGtm5cBPgS+6+zNjb3D3PdHTITP7FvCnU2yDyBm1r2eQ53Yc5PnXO9nTNUh73xAH+oZo7x06/Mu/JBXnmqZaPvvOJt52fj3zKvQLX85+uQTCWqDJzBqB3cAtwIfGlXkIuBW4z8xqCV1I280sBfwY+I67/zD7DWY23933WJicfBOw4dSaIjJ9MhnnYP/w4R397s4BWl7rZO3Og7zW0Q9AcSrOwsoi6soKeFND5eGunjc2VPDmpdWkEjqRSmaWSQPB3dNm9mlgDWF84F5332hmdwIt7r462najmW0iTC/9vLt3mNmHgbcCNWb20egjx6aXfs/M6ghdUuuBP5juxsnslsk4u7sGiMeMquIUhcnYCU9+GjvrdtOenujM2x72dg+ScXAcd3CHoXSGzv5hRsedhVtVnOTNS6v5/auWsKKxmovmlx++MqbIbGDuM6dbvrm52VtaWvJdDTnDMhmn49AwbV0DvLynh017etjUFnboh7IGZ1OJGJVFSSqLk8RjYQpmejTDSHT54I5DR3byRck4y+eVsai6mLiBmWEGhpGMG7WlBdSWpqgrK6S2NMU55YUsqSnW2bYyI5nZOndvnqyczlSWM8bdD1+lsi+6umTvUDqcNJV1meLugRE6+0fY3zvE/p5B9vcOHfVrvbQgwUXzy/md5kVcMC/cMKRrYISu/hG6+oejX/eQShiJWHTnqViMurICLpxfzoXzy1iis25FjqFAkNOiZ3CEDbu72by3l1f39fLK3l5e3dt71C/6iRQmY1QUJaksSlFfXkBTfS3nlBcwr7yQc8oLw6/6quJZfxlikXxQIMi02tjWzbd/tZOH17cxlA4zjKuKkyyfV8bNVzSwqLqYssIEpQXJ8FiYoLwwQXlRkoqipKZfiuSRAkFO2chohp9v2Mu3f7WTltc6KUrG+cAVDbz7knksn1dGXWmB+t5FZgAFghxje3sf9z61g8dfaae2rIClNcUsqS5mcU0JCyoK2d87xM6OQ+w8cIgdHf1sb++jdzDN4upivvibF/I7VyyioliXVhCZaRQIAoQB36e3d/DNJ3fwy1f2k0rEePvyevqG0jz/eif/78U2smdhmsGCiiKW1hbz7960gHdeWM9159erb19kBlMgzDHuzr6eIfaM3dYwuqft09s62LSnh5qSFJ99RxMfvmoJdWUFh983nM6wuyvcMKW+rIBFukyyyKyjQJgjuvtH+OG6XfzDM6+xMzrTdkxhMsZ59aXc9dtv4KbLFk64o08lYjTWltBYW3KmqiwiZ5gCYRYYGc2wu3MAJ8zRLytMUJAIZ+1ubOvmu0+/xkPrdzM4kqF5SRUf/Y2lLKouZn5FEQsqC6koSmrQV0QUCDNFJuOHB3Nf7+hn24E+trcfYlt7H6939B9zs/Nk3CgpSNDVP0JRMs77L1vIh69awsULKvLUAhE52ykQzlK9gyM8smEv/7xpHzsOHOL1g/2H5/UDpOIxltQU01RfyrsunkdjbQnJuNE7mKY3Ohu4d3CEZbWlfODyBs36EZFJKRDOIunRDE9uOcCPXtjNo5v2MjiSoaEqXEP/uuV1LK4pYUl1MUtqillYWaQLq4nItFIgnAW27u/jB2tf58cv7OZA3zCVxUluvqKB91/WwOWLK9W/LyJnhAIhTwaGR/nZS3t4YO3rrN3ZSSJmvOPCej5weQPXLa/XtfRF5IxTIJwm+3sGeXLLAZ7c0k5b1+Dh6+1n3HHCUUHvYJrG2hJuf/cFfODyhqPm/YuInGkKhGm07rWDrNm4j399tZ1X9vYCUFua4ty6UuIWIxYL19s3gxsvmsfNVzRw1bJqdQmJyFlBgTANft3axX9fs5kntxwgGTeal1TzhZUXcG1TLRfNL9flHERkRlAgnIKt+3v5H794lUc27KWqOMkXf/NCbl2xmJIC/WcVkZlHe66TlMk4a3ce5Actu3johd0UJeN89h1NfOLaRsoKNddfRGYuBUIO3J1Ne3pYvb6N1S+2sad7kOJUnI9d3cgfXX8e1SWpfFdRROSU5RQIZrYS+BoQB77h7ndNUOaDwJcAB1509w9F628DvhgV+0t3/3a0/grgPqAI+BnwWXd3zhLuzku7u/nFxn08smEP29oPkYgZ1y2v4473XMg7L6ynOKU8FZHZY9I9mpnFgbuBG4BWYK2ZrXb3TVmUeOFnAAAG3ElEQVRlmoA7gKvdvdPM6qP11cCfA82EoFgXvbcT+DqwCniGEAgrgUems3EnK5NxfrWtgzUb9/Lopn3s7RkkZrCisZqPXdPIey6ZT5WOBkRklsrlJ+4KYKu7bwcwsweA9wGbssp8Erg72tHj7vuj9e8CHnX3g9F7HwVWmtkTQLm7Px2t/w5wE3kMhO6BEf74gRd4fHM7hckYbzu/js9ftJy3X1CvEBCROSGXQFgI7Mp63QpcOa7M+QBm9hShW+lL7v7z47x3YbS0TrA+L17d18uq77TQ2jnAn/3WRdy6YjFFKd38RUTmllwCYaJJ9OP7+hNAE3Ad0AA8aWaXnOC9uXxm+HKzVYSuJRYvXpxDdU/Oz17aw5/+8EWKUwnuX3UVb15aPe3fISIyE+RywZxWYFHW6wagbYIyD7v7iLvvADYTAuJ4722Nnp/oMwFw93vcvdndm+vq6nKobm5GM85f/fwV/uh7z7N8Xhk/+cw1CgMRmdNyCYS1QJOZNZpZCrgFWD2uzEPA9QBmVkvoQtoOrAFuNLMqM6sCbgTWuPseoNfMrrJw3YaPAA9PS4ty0D0wwsfuW8vXn9jGrSsW88Cqq5hXUXimvl5E5Kw0aZeRu6fN7NOEnXscuNfdN5rZnUCLu6/myI5/EzAKfN7dOwDM7C8IoQJw59gAM/CHHJl2+ghnaEB5e3sfn/hOC6939PPl91/C71255Ex8rYjIWc/Ooqn/k2pubvaWlpYpv/9fX23n099/nnjM+PqHr+CqZTXTWDsRkbOTma1z9+bJys2JM6vcnW89tZO//Okmzj+njL//SDOLqovzXS0RkbPKrA8Ed+c///gl7n9uFzdedA5/+7uXUqqLz4mIHGPW7xnNjHPrSvnM28/jT955vi5FLSJyHLM+EAA+ce2yfFdBROSspxv3iogIoEAQEZGIAkFERAAFgoiIRBQIIiICKBBERCSiQBAREUCBICIikRl1cTszawdem+Lba4ED01idfJtN7ZlNbYHZ1Z7Z1BaYu+1Z4u6T3lBmRgXCqTCzllyu9jdTzKb2zKa2wOxqz2xqC6g9k1GXkYiIAAoEERGJzKVAuCffFZhms6k9s6ktMLvaM5vaAmrPCc2ZMQQRETmxuXSEICIiJzAnAsHMVprZZjPbama357s+J8PM7jWz/Wa2IWtdtZk9amZboseqfNbxZJjZIjN73MxeNrONZvbZaP2Ma5OZFZrZc2b2YtSW/xqtbzSzZ6O2/MDMUvmu68kws7iZvWBmP4lez8j2mNlOM3vJzNabWUu0bsb9nY0xs0oze9DMXon+/bxlutsz6wPBzOLA3cC7gYuAW83sovzW6qTcB6wct+524Jfu3gT8Mno9U6SB/+juFwJXAZ+K/n/MxDYNAW939zcBlwIrzewq4K+Ar0Rt6QQ+nsc6TsVngZezXs/k9lzv7pdmTc2ciX9nY74G/NzdLwDeRPh/NL3tcfdZvQBvAdZkvb4DuCPf9TrJNiwFNmS93gzMj57PBzbnu46n0LaHgRtmepuAYuB54ErCiUKJaP1Rf39n+wI0RDuWtwM/AWymtgfYCdSOWzcj/86AcmAH0bjv6WrPrD9CABYCu7Jet0brZrJz3H0PQPRYn+f6TImZLQUuA55lhrYp6l5ZD+wHHgW2AV3uno6KzLS/t68C/wnIRK9rmLntceAXZrbOzFZF62bk3xmwDGgHvhV1533DzEqY5vbMhUCwCdZpalWemVkp8E/AH7t7T77rM1XuPurulxJ+Wa8ALpyo2Jmt1dSY2W8B+919XfbqCYrOiPYAV7v75YTu4k+Z2VvzXaFTkAAuB77u7pcBhzgN3V1zIRBagUVZrxuAtjzVZbrsM7P5ANHj/jzX56SYWZIQBt9z9x9Fq2d0m9y9C3iCMC5SaWaJaNNM+nu7Gnivme0EHiB0G32VGdoed2+LHvcDPyYE9kz9O2sFWt392ej1g4SAmNb2zIVAWAs0RTMlUsAtwOo81+lUrQZui57fRuiHnxHMzIBvAi+7+99mbZpxbTKzOjOrjJ4XAe8kDPQ9DtwcFZsRbQFw9zvcvcHdlxL+nTzm7r/HDGyPmZWYWdnYc+BGYAMz8O8MwN33ArvMbHm06h3AJqa7PfkeLDlDAzLvAV4l9O/+l3zX5yTrfj+wBxgh/Er4OKFf95fAluixOt/1PIn2XEPocvg1sD5a3jMT2wS8EXghassG4M+i9cuA54CtwA+BgnzXdQptuw74yUxtT1TnF6Nl49i/+5n4d5bVpkuBlujv7SGgarrbozOVRUQEmBtdRiIikgMFgoiIAAoEERGJKBBERARQIIiISESBICIigAJBREQiCgQREQHg/wMhU99+qV9yVAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(avg_accuracy1)\n",
    "plt.plot(avg_accuracy2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Rovnako ako v predoslom pripade sa opat pozrieme na priebeh co-trainingu pomocou grafu a statisticky. Opat mozeme na grafe vidiet rovnaky trend."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: 0.5855307640297498\n",
      "Max: 0.6484110885733604\n",
      "Avg: 0.6226729772368716\n",
      "Std: 0.015315989465076459\n"
     ]
    }
   ],
   "source": [
    "print(\"Min: \" + str(np.min(model_accuracy1)))\n",
    "print(\"Max: \" + str(np.max(model_accuracy1)))\n",
    "print(\"Avg: \" + str(np.mean(model_accuracy1)))\n",
    "print(\"Std: \" + str(np.std(model_accuracy1)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: 0.6964164976335362\n",
      "Max: 0.7789046653144016\n",
      "Avg: 0.74412891593419\n",
      "Std: 0.022990284392959685\n"
     ]
    }
   ],
   "source": [
    "print(\"Min: \" + str(np.min(model_accuracy2)))\n",
    "print(\"Max: \" + str(np.max(model_accuracy2)))\n",
    "print(\"Avg: \" + str(np.mean(model_accuracy2)))\n",
    "print(\"Std: \" + str(np.std(model_accuracy2)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: 0.6619337390128465\n",
      "Max: 0.7450980392156863\n",
      "Avg: 0.6961460446247464\n",
      "Std: 0.02084243616165622\n"
     ]
    }
   ],
   "source": [
    "score_combined = []\n",
    "\n",
    "for i in range(30):\n",
    "    score_combined.append(scores[i][2])\n",
    "\n",
    "print(\"Min: \" + str(np.min(score_combined)))\n",
    "print(\"Max: \" + str(np.max(score_combined)))\n",
    "print(\"Avg: \" + str(np.mean(score_combined)))\n",
    "print(\"Std: \" + str(np.std(score_combined)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Co sa tyka skore, v priemere vidime mierne zlepsenia (do 5%, ale zlepsili sa vsetky klasifikatory). Taktiez sa znizil rozdiel medzi min a max."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Co-training s prahovou hodnotou (threshold)\n",
    "\n",
    "Ako sme videli, co-training s iteraciami dosiahol dobre vysledky, ale na ukor jedneho klasifikatora sa zhorsoval druhy. Skusime variantu s thresholdom: Algoritmus bezi, kym ma dostatok dat pripadne vie urcit dostatok nov7ch labelov s urcitou nastavenou istotou"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "def co_training_threshold(L, U, col_names1, col_names2, clf1, clf2, \n",
    "                          label='label', random_state=None, X_test=None, y_test=None, threshold=0.8, \n",
    "                         n=5, p=5):\n",
    "    '''\n",
    "    L - labeled data\n",
    "    U - unlabeled data\n",
    "    col_names1 - columns for clf1\n",
    "    col_names2 - columns for clf2\n",
    "    clf1 - classificator1\n",
    "    clf2 - classificator2\n",
    "    label - name of column in L whit label\n",
    "    threshold - threshold of predictions probability\n",
    "    '''\n",
    "    \n",
    "    pred_1 = []\n",
    "    pred_2 = []\n",
    "    while len(U) > n + p:\n",
    "        # step 1: Use L to train a classifier h1 that considers only the x1 portion of x\n",
    "        clf1 = clf1.fit(L[col_names1],  L[label])\n",
    "        \n",
    "        # step 2: Use L, to train a classifier h2 that considers only the x2 portion of x\n",
    "        clf2 = clf2.fit(L[col_names2], L[label])\n",
    "        \n",
    "        # step 3: Allow hl to label p positive and n negative examples from U\n",
    "        predicted_prob1 = clf1.predict_proba(U[col_names1])\n",
    "        top_positive1 = [x for x in predicted_prob1[:,1].argsort()[-p:] if predicted_prob1[x,1] > threshold] \n",
    "        top_negative1 = [x for x in predicted_prob1[:,0].argsort()[-p:] if predicted_prob1[x,0] > threshold]\n",
    "        \n",
    "        \n",
    "        \n",
    "        # step 4: Allow hl to label p positive and n negative examples from U\n",
    "        predicted_prob2 = clf2.predict_proba(U[col_names2]) \n",
    "        top_positive2 = [x for x in predicted_prob2[:,1].argsort()[-p:] if predicted_prob2[x,1] > threshold]\n",
    "        top_negative2 = [x for x in predicted_prob2[:,0].argsort()[-n:] if predicted_prob2[x,0] > threshold]\n",
    "        \n",
    "        # step 5: Add these self-labeled examples to L, if less n + p labels are added, break \n",
    "        if(np.unique(np.concatenate((top_positive1,top_positive2, top_negative1, top_negative2))).size < n + p):\n",
    "            if (X_test is not None and y_test is not None):\n",
    "                pred = clf1.predict(X_test[col_names1])\n",
    "                pred_1.append(accuracy_score(y_test,pred))\n",
    "\n",
    "                pred = clf2.predict(X_test[col_names2])\n",
    "                pred_2.append(accuracy_score(y_test,pred))\n",
    "                \n",
    "            return [clf1,clf2, pred_1, pred_2]\n",
    "        \n",
    "        \n",
    "        positive_ind = U.iloc[np.unique(np.concatenate((top_positive1,top_positive2))),:].index\n",
    "        negative_ind = U.iloc[np.unique(np.concatenate((top_negative1,top_negative2))),:].index\n",
    "        self_labeled = U.loc[np.unique(np.concatenate((positive_ind,negative_ind))),:]\n",
    "        U.drop(self_labeled.index, inplace=True)\n",
    "        self_labeled.loc[positive_ind, label] = 1\n",
    "        self_labeled.loc[negative_ind, label] = 0\n",
    "        L = pd.concat([L, self_labeled])\n",
    "        \n",
    "        if (X_test is not None and y_test is not None):\n",
    "            pred = clf1.predict(X_test[col_names1])\n",
    "            pred_1.append(accuracy_score(y_test,pred))\n",
    "            \n",
    "            pred = clf2.predict(X_test[col_names2])\n",
    "            pred_2.append(accuracy_score(y_test,pred))\n",
    "    \n",
    "    return [clf1, clf2, pred_1, pred_2]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Opat vyberieme na vstupe 20 zaznamov. Threshold dame velmi prisny, 90% istoty pre vyber prvku:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iteration:  0\n",
      "iteration:  1\n",
      "iteration:  2\n",
      "iteration:  3\n",
      "iteration:  4\n",
      "iteration:  5\n",
      "iteration:  6\n",
      "iteration:  7\n",
      "iteration:  8\n",
      "iteration:  9\n",
      "iteration:  10\n",
      "iteration:  11\n",
      "iteration:  12\n",
      "iteration:  13\n",
      "iteration:  14\n",
      "iteration:  15\n",
      "iteration:  16\n",
      "iteration:  17\n",
      "iteration:  18\n",
      "iteration:  19\n",
      "iteration:  20\n",
      "iteration:  21\n",
      "iteration:  22\n",
      "iteration:  23\n",
      "iteration:  24\n",
      "iteration:  25\n",
      "iteration:  26\n",
      "iteration:  27\n",
      "iteration:  28\n",
      "iteration:  29\n"
     ]
    }
   ],
   "source": [
    "scores_threshold = [[0 for _ in range(3)] for _ in range(30)]\n",
    "\n",
    "for x in range(30):\n",
    "    df_x0 = X_train[y_train == 0].sample(10)\n",
    "    df_x1 = X_train[y_train == 1].sample(10)\n",
    "    \n",
    "    train = pd.concat([df_x0,df_x1])\n",
    "    x_train = X_train.drop(train.index)\n",
    "    train['label'] = y_train.loc[train.index]\n",
    "\n",
    "    clf1 = LogisticRegression(n_jobs=-1)\n",
    "    clf2 = LogisticRegression(n_jobs=-1)\n",
    "\n",
    "    new_1, new_2, scores_threshold[x][0], scores_threshold[x][1] = co_training_threshold(L=train.copy(),U=x_train.copy(), \n",
    "                               col_names1=col_names1,\n",
    "                               col_names2=col_names2, clf1=clf1, clf2=clf2, \n",
    "                                                                     X_test=X_test, y_test=y_test, \n",
    "                                                                     n=5,p=5,threshold=0.90)\n",
    "    print(\"iteration: \",x)\n",
    "    \n",
    "    cl = CombinedClassifier(new_1,new_2,col_names1,col_names2)\n",
    "    pred = cl.predict(X_test)\n",
    "    scores_threshold[x][2] = accuracy_score(y_test, pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_accuracy1 = []\n",
    "model_accuracy2 = []\n",
    "for j in range(30):\n",
    "    model_accuracy1.append(scores_threshold[j][0][-1])\n",
    "    model_accuracy2.append(scores_threshold[j][1][-1])    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for j in range(30):\n",
    "    plt.plot(scores_threshold[j][0],'b')\n",
    "    plt.plot(scores_threshold[j][1], 'r')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "V grafe sme vykreslili priebehy hodnoty accuracy vsetkych modelov (dokopy 60), vidime podobny trend, ako v pripade predtym, lepi klasifikator sa zhorsi, kym ten horsi sa mierne zlepsuje"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pre lepsie znazornenie sa pozrieme este na jeden konretny priklad dvojice modelov"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x20d9f123c50>]"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(scores_threshold[20][0],'r')\n",
    "plt.plot(scores_threshold[20][1],'g')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pozrime sa na statistiky skore, aby sme ho mohli porovnat s co-trainingom s iteraciami:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: 0.3793103448275862\n",
      "Max: 0.6355645706558486\n",
      "Avg: 0.5580121703853956\n",
      "Std: 0.07916010456358391\n"
     ]
    }
   ],
   "source": [
    "print(\"Min: \" + str(np.min(model_accuracy1)))\n",
    "print(\"Max: \" + str(np.max(model_accuracy1)))\n",
    "print(\"Avg: \" + str(np.mean(model_accuracy1)))\n",
    "print(\"Std: \" + str(np.std(model_accuracy1)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: 0.35767410412440837\n",
      "Max: 0.7552400270453009\n",
      "Avg: 0.6045300878972278\n",
      "Std: 0.11679353172503369\n"
     ]
    }
   ],
   "source": [
    "print(\"Min: \" + str(np.min(model_accuracy2)))\n",
    "print(\"Max: \" + str(np.max(model_accuracy2)))\n",
    "print(\"Avg: \" + str(np.mean(model_accuracy2)))\n",
    "print(\"Std: \" + str(np.std(model_accuracy2)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: 0.37322515212981744\n",
      "Max: 0.651791751183232\n",
      "Avg: 0.5719630380887987\n",
      "Std: 0.0880584733668945\n"
     ]
    }
   ],
   "source": [
    "score_combined = []\n",
    "\n",
    "for i in range(30):\n",
    "    score_combined.append(scores_threshold[i][2])\n",
    "\n",
    "print(\"Min: \" + str(np.min(score_combined)))\n",
    "print(\"Max: \" + str(np.max(score_combined)))\n",
    "print(\"Avg: \" + str(np.mean(score_combined)))\n",
    "print(\"Std: \" + str(np.std(score_combined)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ako vidime, skore je celkom porovnatelne s co-trainingom s iteraciami, avsak dosahuje mierne horsie vysledky. Zda sa, ze je teda lepsie pouzit ten s iteraciami. Skusme vsak este pridat viacero olabovanych dat."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Vacsi pocet vstupnych oznacenych dat"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Skusme opat zmenu parametrov - zvysenie poctu vybranych vzoriek pozitivnej a negativnej triedy:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iteration: 0\n",
      "iteration: 1\n",
      "iteration: 2\n",
      "iteration: 3\n",
      "iteration: 4\n",
      "iteration: 5\n",
      "iteration: 6\n",
      "iteration: 7\n",
      "iteration: 8\n",
      "iteration: 9\n",
      "iteration: 10\n",
      "iteration: 11\n",
      "iteration: 12\n",
      "iteration: 13\n",
      "iteration: 14\n",
      "iteration: 15\n",
      "iteration: 16\n",
      "iteration: 17\n",
      "iteration: 18\n",
      "iteration: 19\n",
      "iteration: 20\n",
      "iteration: 21\n",
      "iteration: 22\n",
      "iteration: 23\n",
      "iteration: 24\n",
      "iteration: 25\n",
      "iteration: 26\n",
      "iteration: 27\n",
      "iteration: 28\n",
      "iteration: 29\n"
     ]
    }
   ],
   "source": [
    "scores_threshold = [[0 for _ in range(3)] for _ in range(30)]\n",
    "\n",
    "for x in range(30):\n",
    "    df_x0 = X_train[y_train == 0].sample(30)\n",
    "    df_x1 = X_train[y_train == 1].sample(30)\n",
    "    \n",
    "    train = pd.concat([df_x0,df_x1])\n",
    "    x_train = X_train.drop(train.index)\n",
    "    train['label'] = y_train.loc[train.index]\n",
    "\n",
    "    clf1 = LogisticRegression(n_jobs=-1)\n",
    "    clf2 = LogisticRegression(n_jobs=-1)\n",
    "\n",
    "    new_1, new_2, scores_threshold[x][0], scores_threshold[x][1] = co_training_threshold(L=train.copy(),U=x_train.copy(), \n",
    "                               col_names1=col_names1,\n",
    "                               col_names2=col_names2, clf1=clf1, clf2=clf2, \n",
    "                                                                     X_test=X_test, y_test=y_test, \n",
    "                                                                     n=5,p=5,threshold=0.90)\n",
    "    print(\"iteration:\",x)\n",
    "    \n",
    "\n",
    "    cl = CombinedClassifier(new_1,new_2,col_names1,col_names2)\n",
    "    pred = cl.predict(X_test)\n",
    "    scores_threshold[x][2] = accuracy_score(y_test, pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_accuracy1 = []\n",
    "model_accuracy2 = []\n",
    "for j in range(30):\n",
    "    model_accuracy1.append(scores_threshold[j][0][-1])\n",
    "    model_accuracy2.append(scores_threshold[j][1][-1])    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for j in range(30):\n",
    "    plt.plot(scores_threshold[j][0],'b')\n",
    "    plt.plot(scores_threshold[j][1], 'r')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Rovnako sme vykreslili priebehy vsetkych modelov. Tuto su uz mensie vykyvy, trendy su jasnejsie: ten lepsi klasifikator sa zhorsuje, ten horsi sa mierne zlepsi\n",
    "\n",
    "Pozrieme sa este na jednu konretnu dvojicu"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x20d9f21cd30>]"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(scores_threshold[20][0],'r')\n",
    "plt.plot(scores_threshold[20][1],'g')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pozrime sa na skore jednotlivych klasifikatorov a nasledne ich kombinacie:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: 0.5422582826233941\n",
      "Max: 0.6430020283975659\n",
      "Avg: 0.6185485688528285\n",
      "Std: 0.02038371043151163\n"
     ]
    }
   ],
   "source": [
    "print(\"Min: \" + str(np.min(model_accuracy1)))\n",
    "print(\"Max: \" + str(np.max(model_accuracy1)))\n",
    "print(\"Avg: \" + str(np.mean(model_accuracy1)))\n",
    "print(\"Std: \" + str(np.std(model_accuracy1)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: 0.6402974983096686\n",
      "Max: 0.7768762677484787\n",
      "Avg: 0.6808429118773945\n",
      "Std: 0.03584968437832753\n"
     ]
    }
   ],
   "source": [
    "print(\"Min: \" + str(np.min(model_accuracy2)))\n",
    "print(\"Max: \" + str(np.max(model_accuracy2)))\n",
    "print(\"Avg: \" + str(np.mean(model_accuracy2)))\n",
    "print(\"Std: \" + str(np.std(model_accuracy2)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: 0.623394185260311\n",
      "Max: 0.7038539553752535\n",
      "Avg: 0.6455262564796034\n",
      "Std: 0.019091164461717942\n"
     ]
    }
   ],
   "source": [
    "score_combined = []\n",
    "\n",
    "for i in range(30):\n",
    "    score_combined.append(scores_threshold[i][2])\n",
    "\n",
    "print(\"Min: \" + str(np.min(score_combined)))\n",
    "print(\"Max: \" + str(np.max(score_combined)))\n",
    "print(\"Avg: \" + str(np.mean(score_combined)))\n",
    "print(\"Std: \" + str(np.std(score_combined)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Opat, rovnako ako pri co-trainingu s iteraciami, vidime, ze nastalo mierne zlepsenie. V priemere vsak co-training s iteraciami pracoval mierne lepsie."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Self-training\n",
    "\n",
    "Self-training je jedna z metod semi-supervised learningu. Kedze je uz implementovany v kniznici, pouzijeme hotove riesenie a len to porovname s co-trainingom."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "from frameworks.SelfLearning import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = ms.train_test_split(X, y, test_size=0.2, random_state=42)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Analogicky ako pri co-trainingu, vypocitame skore pre self-training. Na vstupne mu dame ako aj co-trainingu, 20 olabovanych zaznamov:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "scores_self_training = []\n",
    "for x in range(30):\n",
    "    \n",
    "    df_x0 = X_train[y_train == 0].sample(10)\n",
    "    df_x1 = X_train[y_train == 1].sample(10)\n",
    "    \n",
    "    train = pd.concat([df_x0,df_x1])\n",
    "    y_train_s = y_train.copy()\n",
    "    y_train_s[y_train.index.difference(train.index)] = -1\n",
    "    \n",
    "    \n",
    "    clf = LogisticRegression(n_jobs=-1)\n",
    "    ssmodel = SelfLearningModel(clf)\n",
    "    ssmodel.fit(X_train.as_matrix(), y_train_s.as_matrix())\n",
    "    \n",
    "    y_pred = ssmodel.predict(X_test)\n",
    "    scores_self_training.append(accuracy_score(y_pred,y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pozrime sa na vysledky, ktore nam self-training dava:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: 0.44421906693711966\n",
      "Max: 0.5963488843813387\n",
      "Avg: 0.5379310344827586\n",
      "Std: 0.04052871665959093\n"
     ]
    }
   ],
   "source": [
    "print(\"Min: \" + str(np.min(scores_self_training)))\n",
    "print(\"Max: \" + str(np.max(scores_self_training)))\n",
    "print(\"Avg: \" + str(np.mean(scores_self_training)))\n",
    "print(\"Std: \" + str(np.std(scores_self_training)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Vo vysledkoch vidime velke vykyvy, to zrejme indikuje, ze vysledok velmi zavisi od toho, ake data budu vybrane ako prve, do prvej iteracie."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Self-training s vacsim poctom vstupnych oznacenych dat\n",
    "\n",
    "Skusme teraz dat na vstupe 60 olabovanych dat, co predstavuje priblizne 1% dat:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iteration: 0\n",
      "iteration: 1\n",
      "iteration: 2\n",
      "iteration: 3\n",
      "iteration: 4\n",
      "iteration: 5\n",
      "iteration: 6\n",
      "iteration: 7\n",
      "iteration: 8\n",
      "iteration: 9\n",
      "iteration: 10\n",
      "iteration: 11\n",
      "iteration: 12\n",
      "iteration: 13\n",
      "iteration: 14\n",
      "iteration: 15\n",
      "iteration: 16\n",
      "iteration: 17\n",
      "iteration: 18\n",
      "iteration: 19\n",
      "iteration: 20\n",
      "iteration: 21\n",
      "iteration: 22\n",
      "iteration: 23\n",
      "iteration: 24\n",
      "iteration: 25\n",
      "iteration: 26\n",
      "iteration: 27\n",
      "iteration: 28\n",
      "iteration: 29\n"
     ]
    }
   ],
   "source": [
    "scores_self_training = []\n",
    "for x in range(30):\n",
    "    \n",
    "    print(\"iteration:\",x)\n",
    "    df_x0 = X_train[y_train == 0].sample(30)\n",
    "    df_x1 = X_train[y_train == 1].sample(30)\n",
    "    \n",
    "    train = pd.concat([df_x0,df_x1])\n",
    "    y_train_s = y_train.copy()\n",
    "    y_train_s[y_train.index.difference(train.index)] = -1\n",
    "    \n",
    "    \n",
    "    clf = LogisticRegression(n_jobs=-1)\n",
    "    ssmodel = SelfLearningModel(clf)\n",
    "    ssmodel.fit(X_train.as_matrix(), y_train_s.as_matrix())\n",
    "    \n",
    "    y_pred = ssmodel.predict(X_test)\n",
    "    scores_self_training.append(accuracy_score(y_pred,y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pozrime sa na skore s vyssim poctom olabovanych dat:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: 0.4847870182555781\n",
      "Max: 0.6105476673427992\n",
      "Avg: 0.5751183231913455\n",
      "Std: 0.02580110836571806\n"
     ]
    }
   ],
   "source": [
    "print(\"Min: \" + str(np.min(scores_self_training)))\n",
    "print(\"Max: \" + str(np.max(scores_self_training)))\n",
    "print(\"Avg: \" + str(np.mean(scores_self_training)))\n",
    "print(\"Std: \" + str(np.std(scores_self_training)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predikovanie pomocou maleho poctu oznacenych dat bez co-trainingu\n",
    "\n",
    "Pozrime sa, ci ma vobec co-training zmysel. Skusme poslat rovnako, len 20 olabovanych dat do nasho benchmarkoveho modelu a uvidime, ake dosiahne vysledky."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [],
   "source": [
    "scores_supervised = []\n",
    "\n",
    "for x in range(30):\n",
    "    \n",
    "    df_x0 = X_train[y_train == 0].sample(10)\n",
    "    df_x1 = X_train[y_train == 1].sample(10)\n",
    "    \n",
    "    train = pd.concat([df_x0,df_x1])\n",
    "    train['label'] = y_train.loc[train.index]\n",
    "    \n",
    "    benchmark_clf = RandomForestClassifier(n_estimators=500, max_depth=20,\n",
    "                              random_state=10)\n",
    "    \n",
    "    benchmark_clf.fit(train[train.columns.difference(['label'])], train['label'])\n",
    "    y_pred = benchmark_clf.predict(X_test)\n",
    "    scores_supervised.append(accuracy_score(y_pred,y_test))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pozrime si statistiku uspesnosti predikcie (metrika accuracy):"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: 0.4746450304259635\n",
      "Max: 0.5977011494252874\n",
      "Avg: 0.49878296146044626\n",
      "Std: 0.020560173508884495\n"
     ]
    }
   ],
   "source": [
    "print(\"Min: \" + str(np.min(scores_supervised)))\n",
    "print(\"Max: \" + str(np.max(scores_supervised)))\n",
    "print(\"Avg: \" + str(np.mean(scores_supervised)))\n",
    "print(\"Std: \" + str(np.std(scores_supervised)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ako vidime, accuracy skore je naozaj velmi podobne nahode. **Vzhladom nato, ze co-training dokazal dosiahnut lepsie vysledky, usudzujeme, ze stale ma zmysel ho pouzit.**"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}