set.seed(1234)
#import datasets
green <- read.csv("C:/Users/tiago/Desktop/CDados/dataset2/green.csv")
hinselmann <- read.csv("C:/Users/tiago/Desktop/CDados/dataset2/hinselmann.csv")
schiller <- read.csv("C:/Users/tiago/Desktop/CDados/dataset2/schiller.csv")
group <- rbind(green,hinselmann,schiller)
nb <- NULL


#turn some variables factors
factorize <- function(dataset){
  dataset$experts..0 <- as.factor(dataset$experts..0)
  dataset$experts..1 <- as.factor(dataset$experts..1)
  dataset$experts..2 <- as.factor(dataset$experts..2)
  dataset$experts..3 <- as.factor(dataset$experts..3)
  dataset$experts..4 <- as.factor(dataset$experts..4)
  dataset$experts..5 <- as.factor(dataset$experts..5)
  dataset$consensus <- as.factor(dataset$consensus)

  return(dataset)

}

group <- factorize(group)
green <- factorize(green)
hinselmann <- factorize(hinselmann)
schiller <- factorize(schiller)

#remove cervix_area because we will have problems in preProcess with it in future computations
group$cervix_area <- NULL
green$cervix_area <- NULL
hinselmann$cervix_area <- NULL
schiller$cervix_area <- NULL

# Random splitting of data as 70% train and 30%test datasets

#group
indexC_group <- createDataPartition(group$consensus, p=0.70, list=FALSE)
trainDataC_group = group[indexC_group,]
testDataC_group = group[-indexC_group,]



#green
indexC_green <- createDataPartition(green$consensus, p=0.70, list=FALSE)
trainDataC_green = green[indexC_green,]
testDataC_green = green[-indexC_green,]


#hinselmann
indexC_hinselmann <- createDataPartition(hinselmann$consensus, p=0.70, list=FALSE)
trainDataC_hinselmann = hinselmann[indexC_hinselmann,]
testDataC_hinselmann = hinselmann[-indexC_hinselmann,]


#schiller
indexC_schiller <- createDataPartition(schiller$consensus, p=0.70, list=FALSE)
trainDataC_schiller = schiller[indexC_schiller,]
testDataC_schiller = schiller[-indexC_schiller,]


#clustering
cl <- makeCluster(8, type="SOCK")
registerDoSNOW(cl)

grid <- data.frame(fL=c(0.5,1.0), usekernel = TRUE, adjust=c(0.5,1.0))


# define training control
train_control <- trainControl(method="repeatedcv", number=10)

#train based on consensus with and without pp
nb.model_nbC_group_cv = train(trainDataC_group[1:(length(trainDataC_group)-8)], trainDataC_group$consensus, method="nb", trControl=train_control, tuneGrid=grid)
nb.model_nbC_group_pp_c_S_cv = train(trainDataC_group[1:(length(trainDataC_group)-8)], trainDataC_group$consensus, method="nb", trControl=train_control, tuneGrid=grid, preProcess=c("center", "scale"))
nb.model_nbC_group_pp_pca_cv = train(trainDataC_group[1:(length(trainDataC_group)-8)], trainDataC_group$consensus, method="nb", trControl=train_control, tuneGrid=grid, preProcess=c("center", "scale","pca"))

nb.model_nbC_green_cv = train(trainDataC_green[1:(length(trainDataC_green)-8)], trainDataC_green$consensus, method="nb", trControl=train_control, tuneGrid=grid)
nb.model_nbC_green_pp_c_S_cv = train(trainDataC_green[1:(length(trainDataC_green)-8)], trainDataC_green$consensus, method="nb", trControl=train_control, tuneGrid=grid, preProcess=c("center", "scale"))
nb.model_nbC_green_pp_pca_cv = train(trainDataC_green[1:(length(trainDataC_green)-8)], trainDataC_green$consensus, method="nb", trControl=train_control, tuneGrid=grid, preProcess=c("center", "scale","pca"))

nb.model_nbC_hinselmann_cv = train(trainDataC_hinselmann[1:(length(trainDataC_hinselmann)-8)], trainDataC_hinselmann$consensus, method="nb", trControl=train_control, tuneGrid=grid)
nb.model_nbC_hinselmann_pp_c_S_cv = train(trainDataC_hinselmann[1:(length(trainDataC_hinselmann)-8)], trainDataC_hinselmann$consensus, method="nb", trControl=train_control, tuneGrid=grid, preProcess=c("center", "scale"))
nb.model_nbC_hinselmann_pp_pca_cv = train(trainDataC_hinselmann[1:(length(trainDataC_hinselmann)-8)], trainDataC_hinselmann$consensus, method="nb", trControl=train_control, tuneGrid=grid, preProcess=c("center", "scale","pca"))

nb.model_nbC_schiller_cv = train(trainDataC_schiller[1:(length(trainDataC_schiller)-8)], trainDataC_schiller$consensus, method="nb", trControl=train_control, tuneGrid=grid)
nb.model_nbC_schiller_pp_c_S_cv = train(trainDataC_schiller[1:(length(trainDataC_schiller)-8)], trainDataC_schiller$consensus, method="nb", trControl=train_control, tuneGrid=grid, preProcess=c("center", "scale"))
nb.model_nbC_schiller_pp_pca_cv = train(trainDataC_schiller[1:(length(trainDataC_schiller)-8)], trainDataC_schiller$consensus, method="nb", trControl=train_control, tuneGrid=grid, preProcess=c("center", "scale","pca"))

stopCluster(cl)

nb.model_nbC_group_cv$results$Accuracy
nb.model_nbC_group_pp_c_S_cv$results$Accuracy
nb.model_nbC_group_pp_pca_cv$results$Accuracy

nb.model_nbC_green_cv$results$Accuracy
nb.model_nbC_green_pp_c_S_cv$results$Accuracy
nb.model_nbC_green_pp_pca_cv$results$Accuracy

nb.model_nbC_hinselmann_cv$results$Accuracy
nb.model_nbC_hinselmann_pp_c_S_cv$results$Accuracy
nb.model_nbC_hinselmann_pp_pca_cv$results$Accuracy

nb.model_nbC_schiller_cv$results$Accuracy
nb.model_nbC_schiller_pp_c_S_cv$results$Accuracy
nb.model_nbC_schiller_pp_pca_cv$results$Accuracy