In ShanLu92/FeaAug: SVM with Feature Augmentation

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

Feature Augmentation (FeaAug)

This package aims to improve the accuracy of classification models by transfroming the original features into new features that are of better quality. We will updata the package when more methods become available.

In version 0.1.0, this can be accomplished by the function "LMDRT". The function "LMDRT" implements the logarithm marginal density ratios transformation to reach the goal, which greatly improves the detection capability of an SVM-based detection model in NSL-KDD dataset. Other classifiers can also be employed to replace SVM.

Installation

You can instll the package from Github:

#library(devtools)
#install_github('shanlu92/FeaAug')

Or you can download the package from Github and run the following:

#install.packages('path/FeaAug_0.1.0.tar.gz', repos = NULL, type = 'source')

Example

This is a simple example which shows you how to apply the function "LMDRT".

library(e1071)
library(parallel)
library(FeaAug)
intest = subnum(rawdata, k = 10)[[1]] #rawdata is the NSL-KDD dataset.
test = rawdata[intest,]
training = rawdata[-intest,]
data = LMDRT(training, test, 'class', c('normal', 'attack'), T)
traindata = data$traindata
testdata = data$testdata
head(traindata)[,1:6]  # the first six rows and six columns of the transformed train dataset
head(testdata)[,1:6] # the first six rows and six columns of the transformed test dataset

The transformed data are of better quality than the original data. Next, the transformed data can then be put into any classifier such as SVM. More details can be found in example.R.

Main Reference

Wang H , Gu J , Wang S . An effective intrusion detection framework based on SVM with feature augmentation[J]. Knowledge-Based Systems, 2017, 136(Nov.15):130-139.