{"id":1222,"date":"2023-07-27T05:56:29","date_gmt":"2023-07-27T05:56:29","guid":{"rendered":"https:\/\/statorials.org\/tr\/rde-torbalama\/"},"modified":"2023-07-27T05:56:29","modified_gmt":"2023-07-27T05:56:29","slug":"rde-torbalama","status":"publish","type":"post","link":"https:\/\/statorials.org\/tr\/rde-torbalama\/","title":{"rendered":"R'de torbalama nas\u0131l yap\u0131l\u0131r (ad\u0131m ad\u0131m)"},"content":{"rendered":"

<\/p>\n

\n

Belirli bir veri k\u00fcmesi i\u00e7in karar a\u011fac\u0131<\/a> olu\u015fturdu\u011fumuzda, modeli olu\u015fturmak i\u00e7in yaln\u0131zca tek bir e\u011fitim veri k\u00fcmesi kullan\u0131r\u0131z.<\/span><\/p>\n

Bununla birlikte, tek bir karar a\u011fac\u0131 kullanman\u0131n dezavantaj\u0131, y\u00fcksek varyanslara<\/a> maruz kalma e\u011filiminde olmas\u0131d\u0131r. Yani veri setini ikiye b\u00f6l\u00fcp karar a\u011fac\u0131n\u0131 her iki yar\u0131ya da uygularsak sonu\u00e7lar \u00e7ok farkl\u0131 olabilir.<\/span><\/p>\n

Tek bir karar a\u011fac\u0131n\u0131n varyans\u0131n\u0131 azaltmak i\u00e7in kullanabilece\u011fimiz y\u00f6ntemlerden biri, bazen \u00f6ny\u00fckleme toplama<\/em> olarak da adland\u0131r\u0131lan torbalama<\/a> olarak bilinir.<\/span><\/p>\n

Torbalama \u015fu \u015fekilde \u00e7al\u0131\u015f\u0131r:<\/span><\/p>\n

1.<\/strong> Orijinal veri k\u00fcmesinden b<\/em> \u00f6ny\u00fcklemeli \u00f6rnek al\u0131n.<\/span><\/p>\n

2.<\/strong> Her \u00f6ny\u00fckleme \u00f6rne\u011fi i\u00e7in bir karar a\u011fac\u0131 olu\u015fturun.<\/span><\/p>\n

3.<\/strong> Nihai bir model elde etmek i\u00e7in her a\u011fa\u00e7tan elde edilen tahminlerin ortalamas\u0131n\u0131 al\u0131n.<\/span><\/p>\n

Y\u00fczlerce, hatta binlerce bireysel karar a\u011fac\u0131 olu\u015fturarak ve t\u00fcm a\u011fa\u00e7lar\u0131n ortalama tahminlerini alarak, genellikle tek bir karar a\u011fac\u0131yla kar\u015f\u0131la\u015ft\u0131r\u0131ld\u0131\u011f\u0131nda \u00e7ok daha d\u00fc\u015f\u00fck bir test hata oran\u0131 \u00fcreten uygun bir \u00e7anta modeli elde ederiz.<\/span><\/p>\n

Bu e\u011fitimde, R’de paketli bir modelin nas\u0131l olu\u015fturulaca\u011f\u0131na ili\u015fkin ad\u0131m ad\u0131m bir \u00f6rnek sunulmaktad\u0131r.<\/span><\/p>\n

Ad\u0131m 1: Gerekli paketleri y\u00fckleyin<\/strong><\/span><\/h3>\n

\u00d6ncelikle bu \u00f6rnek i\u00e7in gerekli paketleri y\u00fckleyece\u011fiz:<\/span><\/p>\n

 library<\/span> (dplyr) #for data wrangling<\/span>\nlibrary<\/span> (e1071) #for calculating variable importance<\/span>\nlibrary<\/span> (caret) #for general model fitting<\/span>\nlibrary<\/span> (rpart) #for fitting decision trees<\/span>\nlibrary<\/span> (ipred) #for fitting bagged decision trees\n<\/span><\/strong><\/pre>\n
 Ad\u0131m 2: Torbal\u0131 modeli tak\u0131n<\/strong><\/span><\/h3>\n
 Bu \u00f6rnek i\u00e7in, New York City’deki 153 ayr\u0131 g\u00fcn boyunca hava kalitesi \u00f6l\u00e7\u00fcmlerini i\u00e7eren, Hava Kalitesi<\/strong> ad\u0131 verilen yerle\u015fik bir R veri k\u00fcmesini kullanaca\u011f\u0131z.<\/span><\/p>\n
 #view structure of air quality dataset<\/span>\nstr(airquality)\n\n'data.frame': 153 obs. of 6 variables:\n $ Ozone: int 41 36 12 18 NA 28 23 19 8 NA ...\n $Solar.R: int 190 118 149 313 NA NA 299 99 19 194 ...\n $ Wind: num 7.4 8 12.6 11.5 14.3 14.9 8.6 13.8 20.1 8.6 ...\n $ Temp: int 67 72 74 62 56 66 65 59 61 69 ...\n $Month: int 5 5 5 5 5 5 5 5 5 5 ...\n $Day: int 1 2 3 4 5 6 7 8 9 10 ...\n<\/strong><\/pre>\n
 A\u015fa\u011f\u0131daki kod, ipred<\/strong> k\u00fct\u00fcphanesindeki bagging()<\/strong> i\u015flevini kullanarak torbalanm\u0131\u015f bir modelin R’ye nas\u0131l s\u0131\u011fd\u0131r\u0131laca\u011f\u0131n\u0131 g\u00f6sterir.<\/span><\/p>\n
 #make this example reproducible\n<\/span>set.seed(1)\n\n#fit the bagged model\n<\/span>bag <- bagging(\n  formula = Ozone ~ .,\n  data = airquality,\n  nbagg = 150<\/span> ,   \n  coob = TRUE<\/span> ,\n  control = rpart. control<\/span> (minsplit = 2<\/span> , cp = 0<\/span> )\n)\n\n#display fitted bagged model\n<\/span>bag\n\nBagging regression trees with 150 bootstrap replications \n\nCall: bagging.data.frame(formula = Ozone ~ ., data = airquality, nbagg = 150, \n    coob = TRUE, control = rpart.control(minsplit = 2, cp = 0))\n\nOut-of-bag estimate of root mean squared error: 17.4973<\/strong><\/pre>\n
 Torbal\u0131 modeli olu\u015fturmak i\u00e7in 150<\/strong> \u00f6ny\u00fcklemeli \u00f6rnek kullanmay\u0131 se\u00e7ti\u011fimizi ve tahmini torba d\u0131\u015f\u0131 hatas\u0131n\u0131 elde etmek i\u00e7in coob’un<\/strong> DO\u011eRU<\/strong> oldu\u011funu belirtti\u011fimizi unutmay\u0131n.<\/span><\/p>\n
 Ayr\u0131ca rpart.control()<\/strong> i\u015flevinde a\u015fa\u011f\u0131daki \u00f6zellikleri de kulland\u0131k:<\/span><\/p>\n