8 3 2 Destek Vektör Regresyonu Support Vector Regression.ipynb

{
 "metadata": {
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.2-final"
  },
  "orig_nbformat": 2,
  "kernelspec": {
   "name": "Python 3.8.2 64-bit ('.tgy_veribilimi')",
   "display_name": "Python 3.8.2 64-bit ('.tgy_veribilimi')",
   "metadata": {
    "interpreter": {
     "hash": "ba55969ab8b665a0d7ebc430170b2908b5d222710046023b08b5a17088334349"
    }
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2,
 "cells": [
  {
   "source": [
    "# DESTEK VEKTÖR REGRESYONU ( SUPPORT VECTOR REGRESSION )"
   ],
   "cell_type": "markdown",
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np \n",
    "import pandas as pd \n",
    "from sklearn.model_selection import train_test_split, GridSearchCV\n",
    "from sklearn.metrics import mean_squared_error, r2_score\n",
    "import matplotlib.pyplot as plt \n",
    "# Standartlaştırma için eklenen fonksiyonlar scale ve StandartdScaler\n",
    "from sklearn.preprocessing import scale \n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn import model_selection\n",
    "from sklearn.linear_model import LinearRegression\n",
    "from sklearn.tree import DecisionTreeRegressor\n",
    "from sklearn.neighbors import KNeighborsRegressor\n",
    "from sklearn.neural_network import MLPRegressor\n",
    "from sklearn.ensemble import RandomForestRegressor\n",
    "from sklearn.ensemble import GradientBoostingRegressor\n",
    "from sklearn import neighbors\n",
    "from sklearn.svm import SVR"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv(\"verisetleri\\Hitters.csv\")\n",
    "df = df.dropna()\n",
    "dms = pd.get_dummies(df[['League', 'Division', 'NewLeague']])\n",
    "y = df[\"Salary\"]\n",
    "X_ = df.drop(['Salary', 'League', 'Division', 'NewLeague'], axis=1).astype('float64')\n",
    "X = pd.concat([X_, dms[['League_N', 'Division_W', 'NewLeague_N']]], axis=1)\n",
    "X_train, X_test, y_train, y_test =  train_test_split(X, y, test_size=0.25, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# MODEL TAHMİN - Model doğrusal olduğu için \"linear\" değeri verildi. Doğrusal olmayan modle için \"rbf\" (radial bases function) kullanılır"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "svr_model = SVR(\"linear\").fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "SVR(kernel='linear')"
      ]
     },
     "metadata": {},
     "execution_count": 7
    }
   ],
   "source": [
    "svr_model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "array([219.32622169, 702.4303869 , 623.2055964 , 153.77538476,\n",
       "       463.15190834])"
      ]
     },
     "metadata": {},
     "execution_count": 9
    }
   ],
   "source": [
    "svr_model.predict(X_train)[0:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "array([679.14754685, 633.72883069, 925.68640849, 270.28463621,\n",
       "       530.26659184])"
      ]
     },
     "metadata": {},
     "execution_count": 10
    }
   ],
   "source": [
    "svr_model.predict(X_test)[0:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "array([-80.15196151])"
      ]
     },
     "metadata": {},
     "execution_count": 11
    }
   ],
   "source": [
    "svr_model.intercept_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "array([[ -1.21839037,   6.09602969,  -3.67574533,   0.14217075,\n",
       "          0.51435919,   1.28388986,  12.55922537,  -0.08693755,\n",
       "          0.46597184,   2.98259944,   0.52944523,  -0.79820799,\n",
       "         -0.16015534,   0.30872794,   0.28842348,  -1.79560067,\n",
       "          6.41868985, -10.74313783,   1.33374317]])"
      ]
     },
     "metadata": {},
     "execution_count": 12
    }
   ],
   "source": [
    "svr_model.coef_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# test hatası - ilkel olan"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = svr_model.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "370.04084185624924"
      ]
     },
     "metadata": {},
     "execution_count": 16
    }
   ],
   "source": [
    "np.sqrt(mean_squared_error(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Doğrusal olmayanlar için (rbf) için MODEL OLUŞTURMA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "svr_model_rbf = SVR(\"rbf\").fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred_rbf = svr_model_rbf.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "460.0032657244849"
      ]
     },
     "metadata": {},
     "execution_count": 20
    }
   ],
   "source": [
    "np.sqrt(mean_squared_error(y_test, y_pred_rbf))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# MODEL TUNNING \"LINEAR\" İÇİN, C değeri default olarak 1.0 dir."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "svr_params = {\"C\":[0.1, 0.5, 1, 3]}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "svr_cv_model = GridSearchCV(svr_model, svr_params, cv=5).fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "# cv 5 olarak verdik. Çünkü aranacak olan parametre sayısı arttıkça aranacak olan birden fazla parametre ve birden fazla parametrenin birden fazla parmetresi oldukça ve bunlara birde k katlı çapraz doğrulama yöntemi eklendiğinde denenecek olan birçok parametre ortaya çıkmaktadır. Bu sebeple işlemler oldukça uzayabilmektedir. Buraya 10 tane c yazsak, 10 cv yapsak 10-15 dakika sürer. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "{'C': 0.5}"
      ]
     },
     "metadata": {},
     "execution_count": 25
    }
   ],
   "source": [
    "svr_cv_model.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "# GridSearchCV sürelerini görmek için aşağıdaki gibi yazılabilir\n",
    "# verbose = 2 --> Çalışma esnasındaki durumu raporlayarak çalışma gerçekleştirir.\n",
    "# n_jobs = -1 --> Bilgisayarda var olan işlemci gücü max performans ile çalışır."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Fitting 5 folds for each of 4 candidates, totalling 20 fits\n",
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  16 out of  20 | elapsed:   28.4s remaining:    7.0s\n",
      "[Parallel(n_jobs=-1)]: Done  20 out of  20 | elapsed:   55.1s finished\n"
     ]
    }
   ],
   "source": [
    "svr_cv_model = GridSearchCV(svr_model, svr_params, cv=5, verbose=2, n_jobs=-1).fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "{'C': 0.5}"
      ]
     },
     "metadata": {},
     "execution_count": 28
    }
   ],
   "source": [
    "svr_cv_model.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "# FİNAL MODELİ"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "svr_tuned_model = SVR(\"linear\", C = 0.5).fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = svr_tuned_model.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "367.9874739022889"
      ]
     },
     "metadata": {},
     "execution_count": 32
    }
   ],
   "source": [
    "np.sqrt(mean_squared_error(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# son"
   ]
  }
 ]
}