Hi, there: Any body has experience in building a multi-output neural network? <net3<-nnet(train[,c(1:8,10,11)],train[,c(9,12)],size=5, maxit=10000,linout=T) # build a 10-5-2 neural network <pred<-predict(net3, test[,c(1:8,10,11)]) #I am trying to predict two columns of test, which means test[,c(9,12)] I was expecting "pred" be a two-column array. However, it seems the results were mixed up, i.e. the length of "test" data set is 10, instead of getting a 10x2 array of predicting results, I got a 20x1 array. How can I distinguish which of them are predicting results for which column? Thank you very much for helping me out. Qin