library(caret)
library(dplyr)
library(readr)

#Combine 4 data files into one dataframe
Leedata2 <- list.files(path="D:/Work/TaylorsU/Lee Master Thesis/Dataset/Traning and Testing Tets (5% of the entier dataset)/All features", full.names = TRUE) %>% 
  lapply(read_csv) %>% 
  bind_rows 

#Extract 10 features
Leedata3 <- select(Leedata2, c(flgs_number,srate,drate,rate,max,state_number,mean,min,stddev,seq,category))

#Category change to factor for classification
Leedata3$category <- as.factor(Leedata3$category)

#Static summary of dataframe
summary(Leedata3)
#Get data type of dataframe
apply(Leedata3, class)

#Extract 100 DDoS, 100 Dos, 100 Reconnaissance, and 76 Theft data
LeeDDos <- subset(Leedata3,category=="DDoS")
LeeDDos <- sample_n(LeeDDos,100)

LeeDos <- subset(Leedata3,category=="DoS")
LeeDos <- sample_n(LeeDos,100)

LeeRecon <- subset(Leedata3,category=="Reconnaissance")
LeeRecon <- sample_n(LeeRecon,100)

LeeTheft <- subset(Leedata3,category=="Theft")
Leedata4 <- rbind(LeeDDos,LeeDos,LeeRecon,LeeTheft)

#Remove normal factor level
Leedata5 <- droplevels(Leedata4)

#export out dataset
write.csv(Leedata5,"D:/Work/TaylorsU/Lee Master Thesis/Dataset/Leedata5.csv",row.names=TRUE)

#split data to 80:20 ratio
LeeTrain <- createDataPartition(Leedata5$category, p=0.8, list=FALSE, times = 1)
Leetraining <- Leedata5[ LeeTrain, ]
Leetesting <- Leedata5[ -LeeTrain, ]

#Random Forest
library("randomForest")
Leemod_RF <- train(category ~ ., data=Leetraining, method="rf")
Leepred_RF = predict(Leemod_RF, newdata=Leetesting)
confusionMatrix(Leepred_RF,Leetesting$category)

#SVM
library("kernlab")
Leemod_SVM <- train(category ~ ., data=Leetraining, method="svmLinear")
Leepred_SVM = predict(Leemod_SVM, newdata=Leetesting)
confusionMatrix(Leepred_SVM,Leetesting$category)