R code for logisitc regression on spam data

##The Spam data

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

n<-nrow(tmp); p<-ncol(tmp);

x<-matrix(0,n,p);

for(j in 1:p) { x[,j]<-tmp[,j]; }

##Split the data according to HTF for comparison

T<-3065;

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

xtr<-x[idx2,];

xts<-x[-idx2,];

mylogit<-glm(xtr[,p] ~ xtr[,1:(p-1)], family=binomial(link="logit"));

b<-mylogit$coefficients;

logits<-matrix(0,nrow(xts),1);

for (i in 1:nrow(xts))

{

logits[i]<-b[1]+sum(xts[i,1:(p-1)]*b[2:p]);

}

logits<-exp(logits)/(1+exp(logits));

classDF <- data.frame(response = xts[,p], predicted = round(logits,0))

z<-xtabs(~ predicted + response, data = classDF)

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

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