Support Vector Machine and Principal Component Analysis for Intrusion Detection System
Published in i-managers Journal on Software Engineering, 2020
Recommended citation: Mukeri, A., and Gaikwad, D. P. (2020). Support Vector Machine and Principal Component Analysis for Intrusion Detection System. i-managers Journal on Software Engineering , 14(3), 42-49 https://imanagerpublications.com/article/17592/
In this paper, an intelligent Network Intrusion Detection is implemented using Support Vector Machine Classifier. NSL KDD dataset is used for training and separate test data to evaluate the performance of the trained model. Different hyper parameter of Support Vector Machine viz. Y and C are used to tuned the model. The performance of this classifier on principal component analysis transformed dataset as well non-transformed dataset is studied and compared. The experimental results show that support vector machine trained on transformed dataset using Principal Component Analysis exhibits 2% less accuracy as compared with classifier trained on non-transferred dataset. However, classifier trained on transformed dataset using Principal Component Analysis take 15% less training time as compared to classifier trained with non-transferred dataset. The lesser accuracy of the Principal Component Analysis transformed data could be interpreted from the explanation of the variance obtained for top Principal Components as they do not capture the linear separation clearly between the two classes.