IMPROVE THE ACCURACY OF CLASSIFIERS PERFORMANCE USING MACHINE LEARNING & DATA PREPROCESSED METHODS ON NSL-KDD DATA SETS

Abstract

Classification is the method of discovering a set of models that describes data classes for the purpose of being able to utilize the model to forecast the class of objects whose class label is unknown. The derived model is based on the analysis of a set of training data set. Since the class label of each training sample provides this step is referred as supervised learning. The manuscript describes a system that uses Feature Selection [17, 18] as a data pre-processing activities. Feature selection may present us with the means to reduce the number of network parameters made while still maintaining or even elevating the accuracy and reducing false negative rates. In this document, we used variance-Threshold method which finds an optimum feature subset that enhances the classification accuracy. After this step various classifiers are used such as support vector machine, KNN [12]. Experiments were conducted on the NSL_KDD dataset to assess the effectiveness of our approach. The results prove that SVM Ranking with variance threshold feature selection approach leads to promising step up to feature selection and enhances classification accuracy. Based on the system output the Accuracy and error rate of each classifier is computed.