{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "ae4ae7de",
   "metadata": {},
   "source": [
    "1. unik dat v praxi"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "3d818740",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of records: 17379, number of columns: 17\n",
      "Features: ['instant', 'dteday', 'season', 'yr', 'mnth', 'hr', 'holiday', 'weekday', 'workingday', 'weathersit', 'temp', 'atemp', 'hum', 'windspeed', 'casual', 'registered', 'cnt']\n",
      "Number of records: 17379, number of features: 14\n",
      "Features: ['season', 'yr', 'mnth', 'hr', 'holiday', 'weekday', 'workingday', 'weathersit', 'temp', 'atemp', 'hum', 'windspeed', 'casual', 'registered']\n",
      "R2  = 1.00\n",
      "MAE = 0.00 rentals per hour\n"
     ]
    }
   ],
   "source": [
    "from ucimlrepo import fetch_ucirepo\n",
    "from sklearn.linear_model import LinearRegression\n",
    "from sklearn.metrics import mean_absolute_error, r2_score\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# load the original dataset\n",
    "bike = fetch_ucirepo(id=275)\n",
    "df_bike = bike.data.original\n",
    "\n",
    "print(f\"Number of records: {len(df_bike)}, number of columns: {df_bike.shape[1]}\")\n",
    "print(\"Features:\", list(df_bike.columns))\n",
    "\n",
    "# remove the target variable (cnt) and two features (instant, dteday)\n",
    "y_bike = df_bike[\"cnt\"]\n",
    "X_bike = df_bike.drop(columns=[\"cnt\", \"instant\", \"dteday\"])\n",
    "\n",
    "print(f\"Number of records: {len(df_bike)}, number of features: {X_bike.shape[1]}\")\n",
    "print(\"Features:\", list(X_bike.columns))\n",
    "\n",
    "X_bike_train, X_bike_test, y_bike_train, y_bike_test = train_test_split(X_bike, y_bike, test_size=0.2, random_state=42)\n",
    "\n",
    "bike_model = LinearRegression().fit(X_bike_train, y_bike_train)\n",
    "y_bike_pred = bike_model.predict(X_bike_test)\n",
    "\n",
    "print(f\"R2  = {r2_score(y_bike_test, y_bike_pred):.2f}\")\n",
    "print(f\"MAE = {mean_absolute_error(y_bike_test, y_bike_pred):.2f} rentals per hour\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "13b46fdd",
   "metadata": {},
   "source": [
    "ukaz prvych 5 riadkov "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "31d38d7d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "   casual  registered  cnt\n",
      "0       3          13   16\n",
      "1       8          32   40\n",
      "2       5          27   32\n",
      "3       3          10   13\n",
      "4       0           1    1\n",
      "True\n"
     ]
    }
   ],
   "source": [
    "# ukážka prvých 5 riadkov – len tieto 3 stĺpce\n",
    "print(df_bike[[\"casual\", \"registered\", \"cnt\"]].head())\n",
    "\n",
    "# platí casual + registered = cnt vo všetkých riadkoch?\n",
    "print((df_bike[\"casual\"] + df_bike[\"registered\"] == df_bike[\"cnt\"]).all())\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a456cbb8",
   "metadata": {},
   "source": [
    "Korelacia"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "6d2da75b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "registered    0.972151\n",
      "casual        0.694564\n",
      "temp          0.404772\n",
      "atemp         0.400929\n",
      "hr            0.394071\n",
      "hum           0.322911\n",
      "yr            0.250495\n",
      "season        0.178056\n",
      "weathersit    0.142426\n",
      "mnth          0.120638\n",
      "windspeed     0.093234\n",
      "holiday       0.030927\n",
      "workingday    0.030284\n",
      "weekday       0.026900\n",
      "Name: cnt, dtype: float64\n"
     ]
    }
   ],
   "source": [
    "# korelácia každého stĺpca s cnt (bez identifikátora a dátumu)\n",
    "corr = df_bike.drop(columns=[\"instant\", \"dteday\"]).corr()[\"cnt\"].drop(\"cnt\")\n",
    "\n",
    "# zoradenie podľa sily vzťahu (abs = ignorujeme znamienko)\n",
    "print(corr.abs().sort_values(ascending=False))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cc2cde66",
   "metadata": {},
   "source": [
    "Permutation Importance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "d9328fdc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "registered    1.393618\n",
      "casual        0.145719\n",
      "mnth          0.000000\n",
      "hr            0.000000\n",
      "season        0.000000\n",
      "yr            0.000000\n",
      "weekday       0.000000\n",
      "holiday       0.000000\n",
      "workingday    0.000000\n",
      "weathersit    0.000000\n",
      "atemp         0.000000\n",
      "temp          0.000000\n",
      "windspeed     0.000000\n",
      "hum           0.000000\n",
      "dtype: float64\n"
     ]
    }
   ],
   "source": [
    "from sklearn.inspection import permutation_importance\n",
    "import pandas as pd\n",
    "\n",
    "result = permutation_importance(bike_model, X_bike_test, y_bike_test,\n",
    "                                scoring=\"r2\", n_repeats=5, random_state=42)\n",
    "\n",
    "importance = pd.Series(result.importances_mean, index=X_bike.columns)\n",
    "print(importance.sort_values(ascending=False))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "76ad84e5",
   "metadata": {},
   "source": [
    "Overenie"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "f881ea2d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True\n"
     ]
    }
   ],
   "source": [
    "print((df_bike[\"casual\"] + df_bike[\"registered\"] == df_bike[\"cnt\"]).all())\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4fde3824",
   "metadata": {},
   "source": [
    "1.1c: odstranenie a novy model, model bez unikajucich dat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "0c868006",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "R2  = 0.39\n",
      "MAE = 104.80 rentals per hour\n"
     ]
    }
   ],
   "source": [
    "# odstránime unikajúce príznaky\n",
    "X_bike_clean = X_bike.drop(columns=[\"casual\", \"registered\"])\n",
    "\n",
    "# rovnaké rozdelenie ako predtým (random_state=42)\n",
    "X_train_c, X_test_c, y_train_c, y_test_c = train_test_split(\n",
    "    X_bike_clean, y_bike, test_size=0.2, random_state=42)\n",
    "\n",
    "clean_model = LinearRegression().fit(X_train_c, y_train_c)\n",
    "y_pred_c = clean_model.predict(X_test_c)\n",
    "\n",
    "print(f\"R2  = {r2_score(y_test_c, y_pred_c):.2f}\")\n",
    "print(f\"MAE = {mean_absolute_error(y_test_c, y_pred_c):.2f} rentals per hour\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8223906d",
   "metadata": {},
   "source": [
    "| Model | R2 | MAE |\n",
    "|---|---|---|\n",
    "| s `casual` a `registered` | 1.00 | 0.00 |\n",
    "| bez nich | 0.39 | 104.80 |\n",
    "\n",
    "Po odstránení unikajúcich príznakov výkon výrazne klesol – model sa v priemere mýli\n",
    "o ~105 výpožičiek za hodinu. Toto je však **reálny** odhad, ako by model fungoval v praxi.\n",
    "Pôvodný výsledok bol falošne dokonalý, lebo model len sčítaval dva stĺpce.\n",
    "Nízke R2 je čiastočne aj dôsledok toho, že `LinearRegression` nezachytí nelineárne\n",
    "vzťahy (napr. špičky ráno a poobede podľa hodiny).\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "pok-isi (3.14.5)",
   "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": "3.14.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}
