R code for Neural network on Spam data

##Uncomment if use this package for the first time

##install.packages("nnet");

library(nnet);

##The spam data

x<-read.table("spam.Data",sep=",");

n<-nrow(x);

p<-ncol(x);

x[,p]<-as.factor(x[,p]);

colnames(x)[p]<-"class";

##Split the data according to HTF for comparison

T<-3065;

idx2<-sample(1:n,T, replace=FALSE);

spam_train<-x[idx2,1:(p-1)];

spam_test<-x[-idx2,1:(p-1)];

labels<-class.ind(x[,p]);

myspam<-nnet(spam_train, labels[idx2,],

size=3, rang=0.1,

decay=5e-4, maxit=200);

summary(myspam);

test.cl <- function(true, pred) {

true <- max.col(true)

cres <- max.col(pred)

table(true, cres)

}

conf<-test.cl(labels[-idx2,], predict(myspam, spam_test));

acc<-sum(diag(conf))/sum(conf);

cat("The accuracy on the test set is", acc,"\n");