```{r}
#2.5        ####################################################################################
library(MASS)
library(e1071)
library("tree")
fit_NB = naiveBayes(as.factor(good_bad) ~., data = train)
fit_T = tree(formula=as.factor(good_bad)~., data=train, split = "deviance")
fit_T= prune.tree(fit_T, 4)


Yfit_NB = predict(fit_NB, newdata = test, type = "raw") # raw ger sannolikheter.
Yfit_T = predict(fit_T,  newdata = test)
scores_NB = matrix(numeric(((2*length(n)))),ncol =2) #TPR and FPR for Naive Bayes
colnames(scores_NB) = c("TPR", "FPR")

scores_T = matrix(numeric(2*length(n)),ncol =2) #TPR and FPR for Tree model
colnames(scores_T) = c("TPR", "FPR")

n= seq(0.05,0.95,0.05)
for (i in n){
index = match(i,n)

Y_hat_NB = ifelse(Yfit_NB[,2]>i,"good","bad")
Y_hat_T =  ifelse(Yfit_T[,2]>i,"good","bad")

CT_NB = table(test$good_bad,Y_hat_NB)
CT_T = table(test$good_bad,Y_hat_T)

#TPR (True Positive Rate) = probability target being true and model predicting true.
#FPR (False Positive Rate) = probability of the target being false while the model predicts it to be true

scores_NB[index,1] = (CT_NB[1,1]/sum(CT_NB[1,])) # TPR = (true|true)/(true|true+false|true)
scores_NB[index,2] = (CT_NB[2,1]/sum(CT_NB[2,])) # FPR = (true|false)/(false|false+true|false)

scores_T[index,1] = (CT_T[1,1]/sum(CT_T[1,])) # TPR = (true|true)/(true|true+false|true)
scores_T[index,2] = (CT_T[2,1]/sum(CT_T[2,])) # FPR = true|false/(false|false+true|false)

}

par( pty = "s")
plot(scores_NB[,2],scores_NB[,1],type = "l", col = "blue", xlim = c(0,1), ylim = c(0,1),
     xlab = "fpr",
     ylab= "tpr")
lines(scores_T[,2],scores_T[,1], col = "green")

```