Given the churn data for training and test, develop models using different techniques i.e. "Logistic Regression", "Decision Trees", "Random Forests" and "Neural Networks".
· Compare the results using tests like Gini, KS, Rank Ordering, etc. for training and test sets.
· Compare the results across all the technique and report the best technique/model for this data.
· Considering the best model is applied, determine the potential business benefit in next 2 months (consider performance window to be 2 months).
· Document all the steps employed in proper sequence along with R-codes
# Model1 is the predicted probability of Y=1 and output the file into the excel and create the K/S. Please note, sort the bins descending #
bins<-10
attach(mod)
mod1<-mod[order(Model1),]
detach(mod)
cutpoints<-quantile(mod1$Model1,(0:bins)/bins)
binned <-cut(mod1$Model1,cutpoints,include.lowest=TRUE)
x1<-cbind(mod1,binned)
write.csv(x1,"D:/Principal Component/ks.csv")