# This is spam.R file
# It contains some commands that might be helpful for 
# the spam problem
	
# Part A:

library(rpart)
spam.data <- read.table("spam.data", sep=",", na.strings=NA)
dim(spam.data)
names(spam.data) <- c("wfmake", "wfaddress", "wfall", "wf3d", "wfour",
	"wfover", "wfremove", "wfinternet", "wforder", "wfmail", 
	"wfreceive", "wfwill", "wfpeople", "wfreport", "wfaddresses", 
	"wffree", "wfbusiness", "wfemail", "wfyou", "wfcredit", "wfyour", 
	"wffont", "wf000", "wfmoney", "wfhp", "wfhpl", "wfgeorge", "wf650", 
	"wflab", "wflabs", "wftelnet", "wf857", "wfdata", "wf415", "wf85", 
	"wftechnology", "wf1999", "wfparts", "wfpm", "wfdirect", "wfcs", 
	"wfmeeting", "wforiginal", "wfproject", "wfre", "wfedu", "wftable", 
	"wfconference", "cfsc", "cfpar", "cfbrack", "cfexc", "cfdollar", 
	"cfpound", "crlaverage", "crllongest", "crltotal", "spam")

spam.data$spam<-as.factor(spam.data$spam)

# Part B
spam.test<-read.table("spam.test",sep=",")
names(spam.test) <- c("wfmake", "wfaddress", "wfall", "wf3d", "wfour",
	"wfover", "wfremove", "wfinternet", "wforder", "wfmail", 
	"wfreceive", "wfwill", "wfpeople", "wfreport", "wfaddresses", 
	"wffree", "wfbusiness", "wfemail", "wfyou", "wfcredit", "wfyour", 
	"wffont", "wf000", "wfmoney", "wfhp", "wfhpl", "wfgeorge", "wf650", 
	"wflab", "wflabs", "wftelnet", "wf857", "wfdata", "wf415", "wf85", 
	"wftechnology", "wf1999", "wfparts", "wfpm", "wfdirect", "wfcs", 
	"wfmeeting", "wforiginal", "wfproject", "wfre", "wfedu", "wftable", 
	"wfconference", "cfsc", "cfpar", "cfbrack", "cfexc", "cfdollar", 
	"cfpound", "crlaverage", "crllongest", "crltotal", "spam")

spam.test$spam<-as.factor(spam.test$spam)

# Part B
spam.train<-read.table("spam.train",sep=",")
names(spam.train) <- c("wfmake", "wfaddress", "wfall", "wf3d", "wfour",
	"wfover", "wfremove", "wfinternet", "wforder", "wfmail", 
	"wfreceive", "wfwill", "wfpeople", "wfreport", "wfaddresses", 
	"wffree", "wfbusiness", "wfemail", "wfyou", "wfcredit", "wfyour", 
	"wffont", "wf000", "wfmoney", "wfhp", "wfhpl", "wfgeorge", "wf650", 
	"wflab", "wflabs", "wftelnet", "wf857", "wfdata", "wf415", "wf85", 
	"wftechnology", "wf1999", "wfparts", "wfpm", "wfdirect", "wfcs", 
	"wfmeeting", "wforiginal", "wfproject", "wfre", "wfedu", "wftable", 
	"wfconference", "cfsc", "cfpar", "cfbrack", "cfexc", "cfdollar", 
	"cfpound", "crlaverage", "crllongest", "crltotal", "spam")

spam.train$spam<-as.factor(spam.train$spam)

my.control <- rpart.control(cp=0, xval=10)
spamtraintree<-rpart(spam~.,data=spam.train,method="class",control=my.control)

#find optimal cp and prune:
#besttraintree <- prune(spamtraintree, cp=youroptimalcp)

#make predictions using your besttraintree:		
predictions <- predict.rpart(besttraintree, newdata=spam.test, type="class")

#compute totaltesterror, spamtesterror, nonspamtesterror:		
truevalues <- spam.test$spam
len <- length(truevalues)
spamtesterror <- 0
nonspamtesterror <- 0
totaltesterror <- 0
	
for (i in 1:len) {
	if (truevalues[i] == 1 & predictions[i] == 0) {
	   spamtesterror <- spamtesterror + 1
	   totaltesterror <- totaltesterror + 1
        }   
        if (truevalues[i] == 0 & predictions[i] == 1) {
	   nonspamtesterror <- nonspamtesterror + 1
	   totaltesterror <- totaltesterror + 1
        }
}	

totaltesterror <- totaltesterror/len
spamtesterror <- spamtesterror/sum(truevalues == 1)
nonspamtesterror <- nonspamtesterror/sum(truevalues == 0)
totaltesterror 
spamtesterror 
nonspamtesterror 
		
		
# Part C:

my.control <- rpart.control(cp=0, xval=10)
loss1 <- matrix(c(0,2,1,0), ncol=2)
spamtraintree1 <- rpart(spam ~ ., data=spam.train, method="class",
	control=my.control, parms=list(loss=loss1))
#find optimal cp and prune:
#besttraintree1 <- prune(spamtraintree1, cp=youroptimalcp)
#etc.
		
loss2 <- matrix(c(0,10,1,0), ncol=2)
spamtraintree2 <- rpart(spam ~ ., data=spam.train, method="class",
	control=my.control, parms=list(loss=loss2))

#find optimal cp and prune:
#besttraintree2 <- prune(spamtraintree2, cp=youroptimalcp)
#etc.