#		March 19, 2009

# Examples of the RWeka Package

# Weka is a collection of machine learning algorithms.


library(RWeka)
quartz()


####  J48 Decision Trees    version of C4.5 algorithm ######

pairs(iris[,-5])

m1 <- J48(Species ~ ., data = iris) 
m1    		  		  ## print and summary 
summary(m1)	  		  ## calls evaluate_Weka_classifier() 
table(iris$Species, predict(m1))		      ## by hand 
plot(m1)

m1 <- J48(Species ~ ., data = iris[,c(-3)])    ## omit petal.length
m1    		  		  ## print and summary 
table(iris$Species, predict(m1))		      ## by hand 
plot(m1)

m1 <- J48(Species ~ ., data = iris[,c(-3,-4)])    ## omit petal.length
m1    		  		  ## print and summary 
table(iris$Species, predict(m1))		      ## by hand 
plot(m1)

m1 <- J48(Species ~ ., data = iris[,c(-2,-3,-4)])    ## omit petal.length
m1    		  		  ## print and summary 
table(iris$Species, predict(m1))		      ## by hand 
plot(m1)



####   Swiss Bank Notes   #######

swiss = read.table("SwissBankNotes.txt", TRUE, "",skip=19)
swiss = as.data.frame(cbind( swiss, c(rep("Genuine",100),rep("Counterfeit",100)) ))
dimnames(swiss)[[2]][7] = "Status"

pairs(swiss[,-7])

s1 <- J48(status ~ ., data = swiss)    ## omit petal.length
summary(s1)
plot(s1)




#####     SPAM       ##########     39.4% spam


p1 <- J48(class ~ ., data = spam)
plot(p1)

p1 <- J48(class ~ ., data = spam[,c(1:57,59)])
plot(p1)
summary(p1)




##########  Pima Tribe   Female Diabetes  ###########

pima = read.table("pima.indians.diabetes3.txt",TRUE)
pairs(pima[,1:7])
pairs(pima[,1:7],col=pima[,8]+1)

p2 = J48( class ~ ., data=pima[,-8] )
summary(p2)
plot(p2)