Comparative Analysis of Ensemble Learning Methods in Classifying Network Intrusions

Ensemble learning methods as compared to a conventional machine learning method can enhance the classification performance of a network intrusion detection system. The main idea of such methods is to emulate the human nature of gathering and weighing various viewpoints and combining them before making an important decision. However, in selecting the best ensemble learning method, proper assessment is imperative. Hence, this paper provides a comparative study of different ensemble learning methods such as Adaptive Boost, Gradient Boost, Random Forest, Extra Trees, and Logistic Regression based on classification performance and computational cost. The NSL-KDD dataset is selected for this research and is pre-processed to train and test the methods in classifying network intrusions. A comparative analysis of each method is evaluated using the performance metrics, accuracy, recall, precision, F1-score, and computational cost based on training and classification speed. The experimental results show that Extra Trees gives the best performance both in classification performance and computational cost.