Self-tuned Ramp Loss K-Support Vector Classification-Regression

Abstract

This thesis is the extension of a previous work done by Hosseini et al., Hosseini has devised an ramp-loss KSCVR algorithm which classifies dataset based on the principle of anomaly-based intrusion detection techniques. That work used an support vector machine in order to classify the datasets into different classes. The Ramp-KSVCR needs some parameters on the basis of which it classifies the dataset. The tuning of parameter is manual which is time consuming and incorrect tuning can lead to wrong classification which can result in many issues like not allowing access to legitimate traffic . So, in this thesis we have several evolutionary algorithms to find the optimal values for these parameters. The evolutionary algorithms we have used are Genetic algorithm, Simulated Annealing, Mesh adoptive direct search and Random Search. For evaluation purposes, we have used the same dataset used by Hosseini and we were able to get more accuracy as compared the one presented by Hosseini in the previous work.