{"id":3621,"date":"2023-07-16T12:08:13","date_gmt":"2023-07-16T12:08:13","guid":{"rendered":"https:\/\/statorials.org\/tr\/rde-dirsek-yontemi\/"},"modified":"2023-07-16T12:08:13","modified_gmt":"2023-07-16T12:08:13","slug":"rde-dirsek-yontemi","status":"publish","type":"post","link":"https:\/\/statorials.org\/tr\/rde-dirsek-yontemi\/","title":{"rendered":"Optimum k\u00fcmeleri bulmak i\u00e7in r'de dirsek y\u00f6ntemi nas\u0131l kullan\u0131l\u0131r?"},"content":{"rendered":"

<\/p>\n

\n

Makine \u00f6\u011freniminde en s\u0131k kullan\u0131lan k\u00fcmeleme algoritmalar\u0131ndan biri k-ortalama k\u00fcmeleme<\/strong> olarak bilinir.<\/span><\/p>\n

K-ortalama k\u00fcmelemesi, bir veri k\u00fcmesindeki her g\u00f6zlemi K<\/em> k\u00fcmesinden birine yerle\u015ftirdi\u011fimiz bir tekniktir.<\/span><\/p>\n

Nihai ama\u00e7, her k\u00fcme i\u00e7indeki g\u00f6zlemlerin birbirine olduk\u00e7a benzer oldu\u011fu, farkl\u0131 k\u00fcmelerdeki g\u00f6zlemlerin ise birbirinden olduk\u00e7a farkl\u0131 oldu\u011fu K<\/em> k\u00fcmeye sahip olmakt\u0131r.<\/span><\/p>\n

K-ortalamal\u0131 k\u00fcmeleme yaparken ilk ad\u0131m, g\u00f6zlemleri yerle\u015ftirmek istedi\u011fimiz k\u00fcmelerin say\u0131s\u0131 olan K<\/em> i\u00e7in bir de\u011fer se\u00e7mektir.<\/span><\/p>\n

K<\/em> i\u00e7in bir de\u011fer se\u00e7menin en yayg\u0131n yollar\u0131ndan biri, x ekseninde k\u00fcme say\u0131s\u0131 ve y ekseninde karelerin toplam\u0131ndan olu\u015fan bir grafik olu\u015fturmay\u0131 ve ard\u0131ndan K de\u011ferini belirlemeyi i\u00e7eren dirsek y\u00f6ntemi<\/strong> olarak bilinir. olay \u00f6rg\u00fcs\u00fcnde bir “diz” veya d\u00f6n\u00fc\u015f\u00fcn g\u00f6r\u00fcnd\u00fc\u011f\u00fc yer.<\/span><\/p>\n

X ekseninde “dizin” olu\u015ftu\u011fu nokta bize k-ortalamal\u0131 k\u00fcmeleme algoritmas\u0131nda kullan\u0131lacak en uygun k\u00fcme say\u0131s\u0131n\u0131 s\u00f6yler.<\/span><\/p>\n

A\u015fa\u011f\u0131daki \u00f6rnek R’de dirsek y\u00f6nteminin nas\u0131l kullan\u0131laca\u011f\u0131n\u0131 g\u00f6sterir.<\/span><\/p>\n

\u00d6rnek: R’de Dirsek Y\u00f6ntemini Kullanmak<\/strong><\/span><\/h2>\n

Bu \u00f6rnek i\u00e7in, 1973’te her ABD eyaletinde 100.000 ki\u015fi ba\u015f\u0131na cinayet, sald\u0131r\u0131 ve tecav\u00fcz nedeniyle tutuklananlar\u0131n say\u0131s\u0131n\u0131 ve ayr\u0131ca her eyaletin kentsel b\u00f6lgelerde ya\u015fayan n\u00fcfusunun y\u00fczdesini i\u00e7eren, R’de yerle\u015fik USArrests<\/strong> veri k\u00fcmesini kullanaca\u011f\u0131z. alanlar. , UrbanPop.<\/span><\/p>\n

A\u015fa\u011f\u0131daki kod, veri k\u00fcmesinin nas\u0131l y\u00fcklenece\u011fini, eksik de\u011ferleri olan sat\u0131rlar\u0131n nas\u0131l kald\u0131r\u0131laca\u011f\u0131n\u0131 ve veri k\u00fcmesindeki her de\u011fi\u015fkenin ortalamas\u0131 0 ve standart sapmas\u0131 1 olacak \u015fekilde nas\u0131l \u00f6l\u00e7eklenece\u011fini g\u00f6sterir:<\/span><\/p>\n

 #load data<\/span>\ndf <-USArrests\n\n#remove rows with missing values<\/span><\/strong>\ndf <- na. omitted<\/span> (df)\n\n#scale each variable to have a mean of 0 and sd of 1<\/span><\/strong>\ndf <- scale(df)\n\n#view first six rows of dataset<\/span>\nhead(df)\n\n               Murder Assault UrbanPop Rape\nAlabama 1.24256408 0.7828393 -0.5209066 -0.003416473\nAlaska 0.50786248 1.1068225 -1.2117642 2.484202941\nArizona 0.07163341 1.4788032 0.9989801 1.042878388\nArkansas 0.23234938 0.2308680 -1.0735927 -0.184916602\nCalifornia 0.27826823 1.2628144 1.7589234 2.067820292\nColorado 0.02571456 0.3988593 0.8608085 1.864967207\n<\/strong><\/pre>\n
 K-means algoritmas\u0131nda kullan\u0131lacak en uygun k\u00fcme say\u0131s\u0131n\u0131 bulmak i\u00e7in, factoextra<\/strong> paketindeki fviz_nbclust()<\/strong> i\u015flevini kullanarak k\u00fcme say\u0131s\u0131 ile kareler toplam\u0131 toplam\u0131n\u0131n grafi\u011fini olu\u015fturaca\u011f\u0131z:<\/span> <\/p>\n
 library<\/span> (cluster)\nlibrary<\/span> (factoextra)\n\n#create plot of number of clusters vs total within sum of squares<\/span>\nfviz_nbclust(df, kmeans, method = \u201c wss<\/span> \u201d)<\/strong> <\/pre>\n
\"R-b\u00fckme<\/p>\n
 Bu grafikte k = 4 k\u00fcmede bir \u201cdiz\u201d veya b\u00fck\u00fclme oldu\u011fu g\u00f6r\u00fclmektedir. Bu, kareler toplam\u0131ndaki toplam\u0131n istikrar kazanmaya ba\u015flad\u0131\u011f\u0131 noktad\u0131r.<\/span><\/p>\n
 Bu bize k-ortalamalar algoritmas\u0131nda kullan\u0131lacak en uygun k\u00fcme say\u0131s\u0131n\u0131n 4 oldu\u011funu s\u00f6yler.<\/span><\/p>\n
 Not<\/strong> : Daha fazla k\u00fcme kullanarak kareler toplam\u0131nda daha d\u00fc\u015f\u00fck bir toplam elde edebilsek de, muhtemelen e\u011fitim verilerini gere\u011finden fazla s\u0131\u011fd\u0131r\u0131r\u0131z<\/a> ve bu nedenle k-ortalamalar algoritmas\u0131 test verileri \u00fczerinde iyi \u00e7al\u0131\u015fmaz.<\/span><\/p>\n
 K<\/em> 4 i\u00e7in en uygun de\u011feri kullanarak veri k\u00fcmesinde k-means k\u00fcmelemesini ger\u00e7ekle\u015ftirmek i\u00e7in k\u00fcme<\/strong> paketindeki kmeans()<\/strong> i\u015flevini kullanabiliriz:<\/span><\/p>\n
 #make this example reproducible\nset. seeds<\/span> (1)<\/span>\n\n#perform k-means clustering with k = 4 clusters\n<\/span>km <- kmeans(df, centers = 4, nstart = 25)\n\n#view results\n<\/span>km\n\nK-means clustering with 4 clusters of sizes 16, 13, 13, 8\n\nCluster means:\n      Murder Assault UrbanPop Rape\n1 -0.4894375 -0.3826001 0.5758298 -0.26165379\n2 -0.9615407 -1.1066010 -0.9301069 -0.96676331\n3 0.6950701 1.0394414 0.7226370 1.27693964\n4 1.4118898 0.8743346 -0.8145211 0.01927104\n\nVector clustering:\n       Alabama Alaska Arizona Arkansas California Colorado \n             4 3 3 4 3 3 \n   Connecticut Delaware Florida Georgia Hawaii Idaho \n             1 1 3 4 1 2 \n      Illinois Indiana Iowa Kansas Kentucky Louisiana \n             3 1 2 1 2 4 \n         Maine Maryland Massachusetts Michigan Minnesota Mississippi \n             2 3 1 3 2 4 \n      Missouri Montana Nebraska Nevada New Hampshire New Jersey \n             3 2 2 3 2 1 \n    New Mexico New York North Carolina North Dakota Ohio Oklahoma \n             3 3 4 2 1 1 \n        Oregon Pennsylvania Rhode Island South Carolina South Dakota Tennessee \n             1 1 1 4 2 4 \n         Texas Utah Vermont Virginia Washington West Virginia \n             3 1 2 1 1 2 \n     Wisconsin Wyoming \n             2 1 \n\nWithin cluster sum of squares by cluster:\n[1] 16.212213 11.952463 19.922437 8.316061\n (between_SS \/ total_SS = 71.2%)\n\nAvailable components:\n\n[1] \"cluster\" \"centers\" \"totss\" \"withinss\" \"tot.withinss\" \"betweenss\"   \n[7] \"size\" \"iter\" \"ifault\"         \n<\/strong><\/pre>\n
 Sonu\u00e7lardan \u015funu g\u00f6rebiliriz:<\/span><\/p>\n