{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "continental-trading",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split, GridSearchCV\n",
    "from sklearn.metrics import classification_report\n",
    "from sklearn.tree import  DecisionTreeClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "graphic-advertiser",
   "metadata": {},
   "outputs": [],
   "source": [
    "#ucitavanje skupa o perunikama\n",
    "df = pd.read_csv(\"C:/Users/student/Desktop/ipIndustija4/vezbe7/iris.csv\")\n",
    "\n",
    "#izdvajanje skupa sa atributima za predvidjanje\n",
    "featurs = df.columns[:4].tolist()\n",
    "x=df[featurs]\n",
    "\n",
    "#izdvajanje skupa sa ciljnim atributom\n",
    "y=df[\"Species\"]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "comparable-profile",
   "metadata": {},
   "outputs": [],
   "source": [
    "# podela podataka na trening i test skup\n",
    "#70% instanci ce biti u trening skupu, a 30% instanci u test skupu\n",
    "x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.3)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "homeless-holiday",
   "metadata": {},
   "outputs": [],
   "source": [
    "# vrednosti parametara klase DecisionTreeClassifier za unakrsnu validacuju\n",
    "parameters = [{'criterion': ['gini', 'entropy'],\n",
    "               'max_depth': [2, 3, 4, 5],\n",
    "               'min_samples_split':[15, 10, 15],\n",
    "               'min_samples_leaf': [2, 4, 6]\n",
    "               }]\n",
    "\n",
    "#za svaki parametar se zadaje lista vrednosti (napomena: ako se zadaje samo\n",
    "# jedna vrednost za neki parametar, mora se zadati kao element liste, npr.\n",
    "# 'criterion': ['gini'])\n",
    "\n",
    "# moguce je praviti i vise recnika u okviru liste, npr.\n",
    "# [{'criterion': ['gini', 'entropy'],\n",
    "#                'min_samples_split':[15, 10, 15]} ,\n",
    "# {               'min_samples_leaf': [2, 4, 6],\n",
    "#                'max_depth': [2, 3, 4, 5],\n",
    "#                }]\n",
    "# tada ce se praviti modeli na osnovu kombinacija za parametre zadate\n",
    "# u prvom recnika, a zatim modeli na osnovu kombinacija za parametre\n",
    "# zadate u drugom recniku\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "bottom-qualification",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=5, estimator=DecisionTreeClassifier(),\n",
       "             param_grid=[{'criterion': ['gini', 'entropy'],\n",
       "                          'max_depth': [2, 3, 4, 5],\n",
       "                          'min_samples_leaf': [2, 4, 6],\n",
       "                          'min_samples_split': [15, 10, 15]}])"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf = GridSearchCV(DecisionTreeClassifier(), parameters, cv=5)\n",
    "\n",
    "#pravi se model sa optimalnim vrednostima za parametre\n",
    "#Bitno: sa GridSearchCV(...) se pronalaze najbolje vrednosti medju zadatim za parametre,\n",
    "#a pozivom fit(x,y) se pravi model sa najboljim vrednostima za parametre koriscenjem\n",
    "#svih instanci trening skupa\n",
    "    \n",
    "clf.fit(x_train, y_train)\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "ranging-straight",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'criterion': 'entropy',\n",
       " 'max_depth': 3,\n",
       " 'min_samples_leaf': 2,\n",
       " 'min_samples_split': 15}"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "seven-bachelor",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Ocena uspeha po klasifikatorima:\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 2, 'min_samples_leaf': 2, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 2, 'min_samples_leaf': 2, 'min_samples_split': 10}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 2, 'min_samples_leaf': 2, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 2, 'min_samples_leaf': 4, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 2, 'min_samples_leaf': 4, 'min_samples_split': 10}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 2, 'min_samples_leaf': 4, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 2, 'min_samples_leaf': 6, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 2, 'min_samples_leaf': 6, 'min_samples_split': 10}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 2, 'min_samples_leaf': 6, 'min_samples_split': 15}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 3, 'min_samples_leaf': 2, 'min_samples_split': 15}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 3, 'min_samples_leaf': 2, 'min_samples_split': 10}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 3, 'min_samples_leaf': 2, 'min_samples_split': 15}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 3, 'min_samples_leaf': 4, 'min_samples_split': 15}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 3, 'min_samples_leaf': 4, 'min_samples_split': 10}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 3, 'min_samples_leaf': 4, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 3, 'min_samples_leaf': 6, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 3, 'min_samples_leaf': 6, 'min_samples_split': 10}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 3, 'min_samples_leaf': 6, 'min_samples_split': 15}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 4, 'min_samples_leaf': 2, 'min_samples_split': 15}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 4, 'min_samples_leaf': 2, 'min_samples_split': 10}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 4, 'min_samples_leaf': 2, 'min_samples_split': 15}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 4, 'min_samples_leaf': 4, 'min_samples_split': 15}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 4, 'min_samples_leaf': 4, 'min_samples_split': 10}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 4, 'min_samples_leaf': 4, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 4, 'min_samples_leaf': 6, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 4, 'min_samples_leaf': 6, 'min_samples_split': 10}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 4, 'min_samples_leaf': 6, 'min_samples_split': 15}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 5, 'min_samples_leaf': 2, 'min_samples_split': 15}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 5, 'min_samples_leaf': 2, 'min_samples_split': 10}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 5, 'min_samples_leaf': 2, 'min_samples_split': 15}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 5, 'min_samples_leaf': 4, 'min_samples_split': 15}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 5, 'min_samples_leaf': 4, 'min_samples_split': 10}\n",
      "0.952 (+/-0.000) za {'criterion': 'gini', 'max_depth': 5, 'min_samples_leaf': 4, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 5, 'min_samples_leaf': 6, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 5, 'min_samples_leaf': 6, 'min_samples_split': 10}\n",
      "0.914 (+/-0.093) za {'criterion': 'gini', 'max_depth': 5, 'min_samples_leaf': 6, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'entropy', 'max_depth': 2, 'min_samples_leaf': 2, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'entropy', 'max_depth': 2, 'min_samples_leaf': 2, 'min_samples_split': 10}\n",
      "0.914 (+/-0.093) za {'criterion': 'entropy', 'max_depth': 2, 'min_samples_leaf': 2, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'entropy', 'max_depth': 2, 'min_samples_leaf': 4, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'entropy', 'max_depth': 2, 'min_samples_leaf': 4, 'min_samples_split': 10}\n",
      "0.914 (+/-0.093) za {'criterion': 'entropy', 'max_depth': 2, 'min_samples_leaf': 4, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'entropy', 'max_depth': 2, 'min_samples_leaf': 6, 'min_samples_split': 15}\n",
      "0.914 (+/-0.093) za {'criterion': 'entropy', 'max_depth': 2, 'min_samples_leaf': 6, 'min_samples_split': 10}\n",
      "0.914 (+/-0.093) za {'criterion': 'entropy', 'max_depth': 2, 'min_samples_leaf': 6, 'min_samples_split': 15}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 3, 'min_samples_leaf': 2, 'min_samples_split': 15}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 3, 'min_samples_leaf': 2, 'min_samples_split': 10}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 3, 'min_samples_leaf': 2, 'min_samples_split': 15}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 3, 'min_samples_leaf': 4, 'min_samples_split': 15}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 3, 'min_samples_leaf': 4, 'min_samples_split': 10}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 3, 'min_samples_leaf': 4, 'min_samples_split': 15}\n",
      "0.924 (+/-0.114) za {'criterion': 'entropy', 'max_depth': 3, 'min_samples_leaf': 6, 'min_samples_split': 15}\n",
      "0.924 (+/-0.114) za {'criterion': 'entropy', 'max_depth': 3, 'min_samples_leaf': 6, 'min_samples_split': 10}\n",
      "0.924 (+/-0.114) za {'criterion': 'entropy', 'max_depth': 3, 'min_samples_leaf': 6, 'min_samples_split': 15}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 4, 'min_samples_leaf': 2, 'min_samples_split': 15}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 4, 'min_samples_leaf': 2, 'min_samples_split': 10}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 4, 'min_samples_leaf': 2, 'min_samples_split': 15}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 4, 'min_samples_leaf': 4, 'min_samples_split': 15}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 4, 'min_samples_leaf': 4, 'min_samples_split': 10}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 4, 'min_samples_leaf': 4, 'min_samples_split': 15}\n",
      "0.924 (+/-0.114) za {'criterion': 'entropy', 'max_depth': 4, 'min_samples_leaf': 6, 'min_samples_split': 15}\n",
      "0.924 (+/-0.114) za {'criterion': 'entropy', 'max_depth': 4, 'min_samples_leaf': 6, 'min_samples_split': 10}\n",
      "0.924 (+/-0.114) za {'criterion': 'entropy', 'max_depth': 4, 'min_samples_leaf': 6, 'min_samples_split': 15}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 5, 'min_samples_leaf': 2, 'min_samples_split': 15}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 5, 'min_samples_leaf': 2, 'min_samples_split': 10}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 5, 'min_samples_leaf': 2, 'min_samples_split': 15}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 5, 'min_samples_leaf': 4, 'min_samples_split': 15}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 5, 'min_samples_leaf': 4, 'min_samples_split': 10}\n",
      "0.962 (+/-0.038) za {'criterion': 'entropy', 'max_depth': 5, 'min_samples_leaf': 4, 'min_samples_split': 15}\n",
      "0.924 (+/-0.114) za {'criterion': 'entropy', 'max_depth': 5, 'min_samples_leaf': 6, 'min_samples_split': 15}\n",
      "0.924 (+/-0.114) za {'criterion': 'entropy', 'max_depth': 5, 'min_samples_leaf': 6, 'min_samples_split': 10}\n",
      "0.924 (+/-0.114) za {'criterion': 'entropy', 'max_depth': 5, 'min_samples_leaf': 6, 'min_samples_split': 15}\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(\"Ocena uspeha po klasifikatorima:\")\n",
    "    \n",
    "means = clf.cv_results_['mean_test_score']  # za svaku koriscenu kombinaciju\n",
    "                                            # vrednosti za parametre dobija se\n",
    "                                            # srednja vrednost mere za ocenu modela dobijena\n",
    "                                            # na osnovu k test delova\n",
    "\n",
    "stds = clf.cv_results_['std_test_score']    # za svaku koriscenu kombinaciju\n",
    "                                            # vrednosti za parametre dobija se\n",
    "                                            # standardna devijacija vrednosti mere\n",
    "                                            # za ocenu modela dobijena\n",
    "                                            # na osnovu k test delova\n",
    "\n",
    "#za svaku koriscenu kombinaciju vrednosti za parametre se izdvaja izvestaj\n",
    "# (srednja vrednost mere za ocenu modela +/- 2 standardne devijacije)\n",
    "for mean, std, params in zip(means, stds, clf.cv_results_['params']):\n",
    "    print(\"%0.3f (+/-%0.03f) za %s\" % (mean, std * 2, params))\n",
    "print()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "equivalent-diploma",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Izvestaj za test skup:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "      setosa       1.00      1.00      1.00         9\n",
      "  versicolor       1.00      0.95      0.97        20\n",
      "   virginica       0.94      1.00      0.97        16\n",
      "\n",
      "    accuracy                           0.98        45\n",
      "   macro avg       0.98      0.98      0.98        45\n",
      "weighted avg       0.98      0.98      0.98        45\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# model napravljen koriscenjem najboljih parametara se\n",
    "# primenjuje na test skup i izdvaja izvestaj o klasifikaciji\n",
    "# (krajnja ocena modela)\n",
    "print(\"Izvestaj za test skup:\")\n",
    "y_true, y_pred = y_test, clf.predict(x_test)\n",
    "print(classification_report(y_true, y_pred))\n",
    "print()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "aggregate-province",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mera  precision\n",
      "Izvestaj za test skup:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "      setosa       1.00      1.00      1.00         9\n",
      "  versicolor       1.00      0.95      0.97        20\n",
      "   virginica       0.94      1.00      0.97        16\n",
      "\n",
      "    accuracy                           0.98        45\n",
      "   macro avg       0.98      0.98      0.98        45\n",
      "weighted avg       0.98      0.98      0.98        45\n",
      "\n",
      "\n",
      "Mera  f1\n",
      "Izvestaj za test skup:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "      setosa       1.00      1.00      1.00         9\n",
      "  versicolor       1.00      0.95      0.97        20\n",
      "   virginica       0.94      1.00      0.97        16\n",
      "\n",
      "    accuracy                           0.98        45\n",
      "   macro avg       0.98      0.98      0.98        45\n",
      "weighted avg       0.98      0.98      0.98        45\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# promena mere za ocenu modela koje ce se korstiti u unakrsnoj validaciji\n",
    "scores = ['precision', 'f1']\n",
    "\n",
    "for score in scores:\n",
    "    print(\"Mera \", score)\n",
    "\n",
    "    clf = GridSearchCV(DecisionTreeClassifier(), parameters, cv=5, scoring='%s_macro' % score)\n",
    "    clf.fit(x_train, y_train)\n",
    "\n",
    "    print(\"Izvestaj za test skup:\")\n",
    "    y_true, y_pred = y_test, clf.predict(x_test)\n",
    "    print(classification_report(y_true, y_pred))\n",
    "    print()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "experienced-power",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.8.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}