trainData4F <- inputdataTrain[which(inputdataTrain$FLAG==1),]
        trainData4G <- inputdataTrain[which(inputdataTrain$FLAG==0),]
        failNum <- nrow(trainData4F)
        goodNum <- nrow(trainData4G)
        gfRate  <- floor(goodNum/failNum)
       
        predTests <- matrix(0,nrow(inputdataTest),gfRate)
        predlast <- matrix(0,nrow(inputdataTest),gfRate)
        rowIndexs <- 1:nrow(trainData4G)
        set.seed(50)
        for(i in 1:(gfRate-1)){
                train <- NULL
                idxtemp <- sample(1:(goodNum-failNum*(i-1)),failNum,replace=FALSE)
                train <- rbind(trainData4G,trainData4F[rowIndexs[idxtemp],])
#                fit <- randomForest(FLAG~., data=train, distribution="bernoulli", importance=TRUE)
#                glm.model<-glm(FLAG~.,data=train,family=binomial(logit))
#                stepModel<- stepAIC(glm.model,direction="both")
#                rfpred[,i]<- glm.pred <- predict(stepModel,  newdata=inputdataTest,type="response")
                fit <- randomForest(FLAG~., data=train, distribution="bernoulli", importance=TRUE)
                predTests[,i] <- as.numeric(predict(fit,newdata=inputdataTest,type="response"))
                predTests[,i] <- predTests[,i]-1
                rowIndexs <- setdiff(rowIndexs,idxtemp)
        }
        predTests[,gfRate] <- inputdataTest$FLAG
        predlast[,1] <- inputdataTest$FLAG
       
        predlast[,2] <- apply(predTests, 1, mean)
#         for(b in 1:nrow(inputdataTest)){
#                predlast[b,2] <- mean(predTests[b,])
#        }
        #############Train the model for whole data set#################