library(dplyr)
library(caret)
library(randomForest)
library(ModelMetrics)
setwd("F:/pajtoon/lab 11")
load('dataLeukemia.RData')

View(data[1:100,1:100])
names(data)[1:10]
unique(data$Leukemia.class)

set.data = data[data$Leukemia.class == "AML with normal karyotype + other abnormalities" |data$Leukemia.class == "CLL", -c(1,3)]
set.data$Leukemia.class = as.numeric(set.data$Leukemia.class)/6-1
View(set.data[,1:100])
rm(data)
v=1:length(set.data$Leukemia.class)

index.test = sample(v,round(length(set.data$Leukemia.class)/3))
index.train = v[-index.test]

data.train = set.data[index.train,]
data.test = set.data[index.test,]

result = randomForest(x=data.train[,-1], y=as.factor(data.train[,1]), importance = TRUE)
var.imp = result$importance  # View(var.imp)
var.imp100 = row.names(var.imp[order(var.imp[,4], decreasing = TRUE),])[1:100]
View(var.imp[order(var.imp[,4], decreasing = TRUE),])
var.pred.test = predict(result, x=data.test[,var.imp100])
print(auc(as.factor(data.test[,1]), var.pred.test))

result2 = randomForest(x=data.train[,var.imp100], y=as.factor(data.train[,1]), importance = TRUE)
var.pred.test = predict(result2, x=data.test[,var.imp100])
print(auc(as.factor(data.test[,1]), var.pred.test))