riper_fun()

[6기_심재현]

######################################################################

#########################3-3 riper_fun() #############################

######################################################################

riper_fun <- function(){

  library(RWeka)

  input_table <- readline('csv파일을 입력하세요. ex) emp.csv : ')

  table_name <- read.csv(input_table, stringsAsFactors = F, header=T)

  print(colnames(table_name))

  input_header <- readline('컬럼명이 있습니까? ex)T OR F : ')

  input_label <- readline('라벨 변수명을 입력하세요. ex)buy_yn : ') 

  input_rm_num <- readline('배제할 컬럼이 있다면 컬럼 위치번호를 입력하세요. ex) n,n,n ..., 없을 시 0 : ')

  ### 헤더 유무 검사

  if (input_header == 'T'){

    table_name <- read.csv(input_table, stringsAsFactors = T, header=T)  

  }else {

    table_name <- read.csv(input_table, stringsAsFactors = T, header=F)

  }

  #컬럼명으로 컬럼위치 찾기

  input_label_num <- which(colnames(table_name)==input_label)

  ### na값 제거

  table_name <- na.omit(table_name)

  ## 라벨 값 추출

  table_label_col<-table_name[, input_label_num]

  ### 제거할 컬럼 제거

  split_num <- strsplit(input_rm_num, ',')

  split_num <- sort(as.integer(c(unlist(split_num), input_label_num)), decreasing = T)

  if (0 %in% split_num ){

    table_name <- table_name[,-as.integer(input_label_num)]

  }else{

    for(i in split_num){

      table_name <- table_name[,-as.integer(i)]

    }

  }

  table2 <- cbind(table_name, lb=factor(table_label_col))

  print(table2)

  ## 결과

  n <- dim(table_name)[2]

  for ( i in 1:n) {

    table2[,i] <- factor(table2[,i])  

  }

  set.seed(1)  

  train_cnt <- round(0.75*dim(table2)[1])

  train_index <- sample(1:dim(table2)[1],train_cnt,

                        replace=F)

  table_train <- table2[train_index, ] 

  table_test  <- table2[-train_index, ]

  print(table_train)

  print(table_test)

  model <- JRip(lb~., data = table_train)

  summary(model)

  result <- predict(model, table_test)

  library(gmodels)

  test_label_num <- which(colnames(table_test)=="lb")

  CrossTable(result, table_test[ , test_label_num] )

}