library(rpart)
library(rpart.plot)
library(arules)

set.seed(1)

clientrating.df$sum_scoreD = discretize(clientrating.df$sum_score, "frequency", categories=4, labels = c("Bad", "Average", "Ok", "Good"))

flag=sample(1:nrow(clientrating.df), nrow(clientrating.df) / 2, replace = F)

clientLrn = clientrating.df[flag,]
clientTst = clientrating.df[-flag,]

ac = rpart(data = clientLrn, sum_scoreD~gender+age+dis_avg_salary+dis_unemp_rate_95+dis_unemp_rate_96+dis_commit_crimes_95+dis_commit_crimes_96+client_type+card+year_card+acnt_frequency+year_account+loan_status)

prp(ac)

Dados.previsto.com.modelo<-predict(ac,clientLrn)
erros.quadraticos<- (clientLrn$sum_score - Dados.previsto.com.modelo)^2
erros.quadraticos
erro.medio.quadratico <- sum(erros.quadraticos) / length(erros.quadraticos)
(erro.medio<- erro.medio.quadratico^0.5)