#
# R-Code for Handout 5
#
#
misclass = function(fit,y) {
temp <- table(fit,y)
cat("Table of Misclassification\n")
cat("(row = predicted, col = actual)\n")
print(temp)
cat("\n\n")
numcor <- sum(diag(temp))
numinc <- length(y) - numcor
mcr <- numinc/length(y)
cat(paste("Misclassification Rate = ",format(mcr,digits=3)))
cat("\n")
}
#
# Read in Breast Cancer diagnosis .csv file.
#
BreastDiag = read.table(file.choose(),header=T,sep=",")
names(BreastDiag)
head(BreastDiag,5)

str(BreastDiag)
par(mfrow=c(3,2))
plot(Radius~Diagnosis,data=BreastDiag,col="red")
plot(Texture~Diagnosis,data=BreastDiag,col="red")
plot(FracDim~Diagnosis,data=BreastDiag,col="red")
plot(Symmetry~Diagnosis,data=BreastDiag,col="red")
plot(Compactness~Diagnosis,data=BreastDiag,col="red")
plot(Concavity~Diagnosis,data=BreastDiag,col="red")
par(mfrow=c(1,1))

sam = sample(1:569,floor(569*.6666),replace=F)

BCtrain = BreastDiag[sam,]
BCtest = BreastDiag[-sam,]

dim(BCtrain)
dim(BCtest)

bc.rpart = rpart(Diagnosis~.,data=BCtrain)
plot(bc.rpart)
text(bc.rpart)

yfit = predict(bc.rpart,type="class")
misclass(yfit,BCtrain$Diagnosis)

# This command may not work.
post(bc.rpart)

yprob = predict(bc.rpart)
head(yprob,5)
summary(bc.rpart)

bc.rpart2 = rpart(Diagnosis~.,data=BCtrain,cp=.0001,minsplit=4)
yfit = predict(bc.rpart2)
yfit = predict(bc.rpart2,type="class")
misclass(yfit,BCtrain$Diagnosis)

prp(bc.rpart2,type=4,extra=3,cex=.7)

ypred = predict(bc.rpart,newdata=BCtest,type="class")
misclass(ypred,BCtest$Diagnosis)

ypred2 = predict(bc.rpart2,newdata=BCtest,type="class")
misclass(ypred2,BCtest$Diagnosis)

crpart.sscv = function(fit,y,data,B=25,p=.333) {
n = length(y)
cv <- rep(0,B)
for (i in 1:B) {
       ss <- floor(n*p)
       sam <- sample(1:n,ss)
       temp <- data[-sam,]
       fit2 <- rpart(formula(fit),data=temp,parms=fit$parms,control=fit$control)
       ynew <- predict(fit2,newdata=data[sam,],type="class")
       tab <- table(y[sam],ynew)
       mc <- ss - sum(diag(tab))
      cv[i] <- mc/ss
      }
   cv
}

bc.rpart3 = rpart(Diagnosis~.,data=BCtrain,cp=.00001,minsplit=5)
results = crpart.sscv(bc.rpart3,BCtrain$Diagnosis,data=BCtrain,B=200)
summary(results)

plotcp(bc.rpart3)

bc.rpart4 = rpart(Diagnosis~.,data=BCtrain,cp=.019,minsplit=5)
results = crpart.sscv(bc.rpart4,BCtrain$Diagnosis,data=BCtrain,B=200)
summary(results)

ypred = predict(bc.rpart4,newdata=BCtest,type="class")
misclass(ypred,BCtest$Diagnosis)

bc.bag = bagging(Diagnosis~.,data=BCtrain,nbagg=100,coob=T
control=rpart.control(cp=.019,minsplit=5,xval=0))
bc.bag

ypred = predict(bc.bag,newdata=BCtest,type="class")
misclass(ypred,BCtest$Diagnosis)

library(randomForest)
bc.rf = randomForest(Diagnosis~.,data=BCtrain)
bc.rf
varImpPlot(bc.rf)
ypred = predict(bc.rf, newdata=BCtest,type="class")
misclass(ypred,BCtest$Diagnosis)


bc.boost = ada(Diagnosis~.,data=BCtrain)
bc.boost
summary(bc.boost)

ypred = predict(bc.boost,newdata=BCtest)
attributes(ypred)
misclass(ypred,BCtest$Diagnosis)

#
# Task - Classifying Olive Oils - read in the file OliveOils.csv
#
OliveOils = read.table(file.choose(),header=T,sep=",") 
names(OliveOils)
dim(OliveOils)
train = sample(1:572,floor(.6666*572),replace=F)
Olive.train = OliveOils[train,]
Olive.test = OliveOils[-train,]
dim(Olive.train)
dim(Olive.test)
#
# You will have to replace "other settings you choose" with something!!
#
olive.bag = bagging(Area.name~.,data=Olive.train, other settings you choose)
ypred = predict(olive.bag,newdata=Olive.test,type="class")
attributes(ypred)
ypred$error
misclass(ypred$class,Olive.test$Area.name)
#
# Again you have some tunning parameters to fill in.
#
olive.boost = boosting(Area.name~.,data=Olive.train,mfinal=100,
    control=rpart.control(change cp, minsplit, etc. here if you want))
ypred = predict(olive.boost,newdata=Olive.test)
attributes(ypred)
misclass(ypred$class,Olive.test$Area.name)	

#
# Task 2 - Mushrooms - read in the file Mushrooms.csv
#
Mushrooms = read.table(file.choose(),header=T,sep=",")
dim(Mushrooms) 
names(Mushrooms)
summary(Mushrooms)
str(Mushrooms)
Mushrooms2 = Mushrooms[,-c(12,17)]
train = sample(1:nrow(Mushrooms2),floor(nrow(Mushrooms*0.5)),replace=F)
Mush.train = Mushrooms2[train,]
Mush.test = Mushrooms2[-train,]
mush.tree = rpart(Poisonous~.,data=Mush.train)
summary(mush.tree)
misclass(Mush.train$Poisonous,predict(mush.tree,type="class"))
ypred = predict(mush.tree,newdata=Mush.test,type="class")
misclass(Mush.test$Poisonous,ypred)