<\/p>\n
Lojistik regresyon,<\/a> yan\u0131t de\u011fi\u015fkeni ikili oldu\u011funda bir regresyon modeline uymak i\u00e7in kullanabilece\u011fimiz bir y\u00f6ntemdir.<\/span><\/p>\n R’ye bir lojistik regresyon modeli yerle\u015ftirdi\u011finizde, model \u00f6zetindeki katsay\u0131lar, her \u00f6ng\u00f6r\u00fcc\u00fc de\u011fi\u015fkendeki bir birimlik art\u0131\u015fla ili\u015fkili yan\u0131t de\u011fi\u015fkeninin log oranlar\u0131ndaki<\/strong> ortalama de\u011fi\u015fimi temsil eder.<\/span><\/p>\n Ancak bunun yerine s\u0131kl\u0131kla modeldeki yorday\u0131c\u0131 de\u011fi\u015fkenlerin olas\u0131l\u0131k oran\u0131n\u0131<\/strong> hesaplamak isteriz.<\/span><\/p>\n Modeldeki her tahmin de\u011fi\u015fkenine ili\u015fkin olas\u0131l\u0131k oranlar\u0131n\u0131 h\u0131zl\u0131 bir \u015fekilde hesaplamak i\u00e7in a\u015fa\u011f\u0131daki s\u00f6zdizimini kullanabilirsiniz:<\/span><\/p>\n A\u015fa\u011f\u0131daki s\u00f6zdizimini kullanarak her olas\u0131l\u0131k oran\u0131 i\u00e7in %95 g\u00fcven aral\u0131\u011f\u0131n\u0131 da hesaplayabilirsiniz:<\/span><\/p>\n A\u015fa\u011f\u0131daki \u00f6rnek, R’deki bir lojistik regresyon modeli i\u00e7in olas\u0131l\u0131k oranlar\u0131n\u0131 hesaplamak ve yorumlamak amac\u0131yla bu s\u00f6zdiziminin nas\u0131l kullan\u0131laca\u011f\u0131n\u0131 g\u00f6sterir.<\/span><\/p>\n Bu \u00f6rnek i\u00e7in R’deki ISLR<\/strong> paketindeki varsay\u0131lan<\/strong> veri k\u00fcmesini kullanaca\u011f\u0131z.<\/span><\/p>\n Veri k\u00fcmesinin \u00f6zetini y\u00fcklemek ve g\u00f6r\u00fcnt\u00fclemek i\u00e7in a\u015fa\u011f\u0131daki kodu kullanabiliriz:<\/span><\/p>\n Bu veri seti 10.000 ki\u015fiye ili\u015fkin a\u015fa\u011f\u0131daki bilgileri i\u00e7ermektedir:<\/span><\/p>\n Belirli bir bireyin temerr\u00fcde d\u00fc\u015fme olas\u0131l\u0131\u011f\u0131n\u0131 tahmin eden bir lojistik regresyon modeli olu\u015fturmak i\u00e7in \u00f6\u011frenci durumunu, banka bakiyesini ve geliri kullanaca\u011f\u0131z.<\/span><\/p>\n R’nin veri k\u00fcmesine bir lojistik regresyon modeli uydurmas\u0131 i\u00e7in glm<\/strong> i\u015flevini kullanabilir ve family=’binomial’ belirtebiliriz:<\/span><\/p>\n Sonu\u00e7taki katsay\u0131lar, log temerr\u00fct olas\u0131l\u0131klar\u0131ndaki ortalama de\u011fi\u015fimi g\u00f6stermektedir.<\/span><\/p>\n \u00d6rne\u011fin, bakiyedeki<\/strong> bir birimlik art\u0131\u015f, log temerr\u00fct olas\u0131l\u0131\u011f\u0131nda ortalama 0,005988’lik<\/strong> bir art\u0131\u015fla ili\u015fkilidir.<\/span><\/p>\n Bunun yerine her tahmin de\u011fi\u015fkeninin olas\u0131l\u0131k oran\u0131n\u0131 hesaplamak i\u00e7in a\u015fa\u011f\u0131daki s\u00f6zdizimini kullanabiliriz:<\/span><\/p>\n Ayr\u0131ca her bir oran oran\u0131n\u0131 ve her bir oran oran\u0131 i\u00e7in %95 g\u00fcven aral\u0131\u011f\u0131n\u0131 da hesaplayabiliriz:<\/span><\/p>\n Her bir katsay\u0131 i\u00e7in olas\u0131l\u0131k oran\u0131, di\u011fer t\u00fcm \u00f6ng\u00f6r\u00fcc\u00fc de\u011fi\u015fkenlerin sabit kald\u0131\u011f\u0131 varsay\u0131larak, bir bireyin temerr\u00fct olas\u0131l\u0131klar\u0131ndaki ortalama art\u0131\u015f\u0131 temsil eder.<\/span><\/p>\n \u00d6rne\u011fin, yorday\u0131c\u0131 de\u011fi\u015fkenin bakiyesinin<\/strong> olas\u0131l\u0131k oran\u0131 1,0057’dir<\/strong> .<\/span><\/p>\n Bu, \u00f6\u011frenci stat\u00fcs\u00fcn\u00fcn ve gelirinin sabit kald\u0131\u011f\u0131 varsay\u0131ld\u0131\u011f\u0131nda<\/em> , bireyin bakiyesinde ta\u015f\u0131d\u0131\u011f\u0131 her ek dolar i\u00e7in, bireyin kredisini \u00f6deyememe ihtimalinin 1,0057<\/strong> kat artt\u0131\u011f\u0131 anlam\u0131na gelir.<\/span><\/p>\n Di\u011fer yorday\u0131c\u0131 de\u011fi\u015fkenlerin odds oranlar\u0131n\u0131 da ayn\u0131 \u015fekilde yorumlayabiliriz.<\/span><\/p>\n A\u015fa\u011f\u0131daki e\u011fitimlerde R’de di\u011fer ortak g\u00f6revlerin nas\u0131l ger\u00e7ekle\u015ftirilece\u011fi a\u00e7\u0131klanmaktad\u0131r:<\/span><\/p>\n Predict () R’de lojistik regresyon modeliyle nas\u0131l kullan\u0131l\u0131r?<\/a> Lojistik regresyon, yan\u0131t de\u011fi\u015fkeni ikili oldu\u011funda bir regresyon modeline uymak i\u00e7in kullanabilece\u011fimiz bir y\u00f6ntemdir. R’ye bir lojistik regresyon modeli yerle\u015ftirdi\u011finizde, model \u00f6zetindeki katsay\u0131lar, her \u00f6ng\u00f6r\u00fcc\u00fc de\u011fi\u015fkendeki bir birimlik art\u0131\u015fla ili\u015fkili yan\u0131t de\u011fi\u015fkeninin log oranlar\u0131ndaki ortalama de\u011fi\u015fimi temsil eder. Ancak bunun yerine s\u0131kl\u0131kla modeldeki yorday\u0131c\u0131 de\u011fi\u015fkenlerin olas\u0131l\u0131k oran\u0131n\u0131 hesaplamak isteriz. Modeldeki her tahmin de\u011fi\u015fkenine ili\u015fkin olas\u0131l\u0131k […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[11],"tags":[],"class_list":["post-4489","post","type-post","status-publish","format-standard","hentry","category-rehber"],"yoast_head":"\n exp(coef(model))\n<\/strong><\/pre>\n
exp(cbind(Odds_Ratio = coef(model), confint(model)))<\/strong><\/pre>\n
\u00d6rnek: R’de Lojistik Regresyon Modelinde Olas\u0131l\u0131k Oranlar\u0131n\u0131n Hesaplanmas\u0131<\/strong><\/span><\/h2>\n
library<\/span> (ISLR)\n\n#view first five rows of Default dataset\n<\/span>head(Default)\n\n default student balance income\n1 No No 729.5265 44361.625\n2 No Yes 817.1804 12106.135\n3 No No 1073.5492 31767.139\n4 No No 529.2506 35704.494\n5 No No 785.6559 38463.496\n6 No Yes 919.5885 7491.559<\/strong><\/pre>\n\n
#fit logistic regression model<\/span>\nmodel <- glm(default~student+balance+income, family=' binomial<\/span> ', data=Default)\n\n#disable scientific notation for model summary<\/span>\noptions(scipen=999)\n\n#view model summary<\/span>\nsummary(model)\n\nCall:\nglm(formula = default ~ student + balance + income, family = \"binomial\", \n data = train)\n\nDeviance Residuals: \n Min 1Q Median 3Q Max \n-2.5586 -0.1353 -0.0519 -0.0177 3.7973 \n\nCoefficients:\n Estimate Std. Error z value Pr(>|z|) \n(Intercept) -11.478101194 0.623409555 -18.412 <0.0000000000000002 ***\nstudentYes -0.493292438 0.285735949 -1.726 0.0843 . \nbalance 0.005988059 0.000293765 20.384 <0.0000000000000002 ***\nincome 0.000007857 0.000009965 0.788 0.4304 \n---\nSignificant. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1\n\n(Dispersion parameter for binomial family taken to be 1)\n\n Null deviance: 2021.1 on 6963 degrees of freedom\nResidual deviance: 1065.4 on 6960 degrees of freedom\nAIC: 1073.4\n\nNumber of Fisher Scoring iterations: 8\n<\/strong><\/pre>\n #calculate odds ratio for each predictor variable<\/span>\nexp(coef(model))\n\n (Intercept) studentYes balance income \n0.00001903854 0.52373166965 1.00575299051 1.00000303345 \n<\/strong><\/pre>\n #calculate odds ratio and 95% confidence interval for each predictor variable<\/span> \nexp(cbind(Odds_Ratio = coef(model), confint(model)))\n\n Odds_Ratio 2.5% 97.5%\n(Intercept) 0.00001903854 0.000007074481 0.0000487808\nstudentYes 0.52373166965 0.329882707270 0.8334223982\nbalance 1.00575299051 1.005308940686 1.0062238757\nincome 1.00000303345 0.999986952969 1.0000191246\n<\/strong><\/pre>\n Ek kaynaklar<\/strong><\/span><\/h2>\n
R’de lojistik regresyon \u00e7\u0131kt\u0131s\u0131nda Pr(>|z|) nas\u0131l yorumlan\u0131r?<\/a>
R’de lojistik regresyon e\u011frisi nas\u0131l \u00e7izilir<\/a><\/span><\/p>\n","protected":false},"excerpt":{"rendered":"