MODEL VALIDATION

CROSS-VALIDATION

We use 10-fold cross validation to validate our models.

R Code:

f#Cross Validation for full model

library(boot)

cost <- function(Attrition,pi=0)

                              {

                                    mean(abs(Attrition-pi)>0.5)

                               }

HR.glm<-glm(Attrition~.,binomial,data=HR)

(cv.err<-cv.glm(HR,HR.glm,cost,K=10)$delta[1])

[1] 0.1340136

backwd.glm<-glm(Attrition~.,binomial,data=HR)

(cv.err<-cv.glm(HR,backwd.glm,cost,K=10)$delta[1])

[1] 0.129932

The result shows that the model obtained by backward selection is associated with the lowest cross validation error and hence we will be using the full model for our prediction.