Hybrid Optimization Algorithm with Random Forest for Class Balancing in Network Traffic Classification

Abstract

This study delves into the intricate task of network traffic classification employing machine learning algorithms, leveraging the KDD dataset. Initially, a range of algorithms, including KNN, Random Forest, SVM, and logistic regression, were explored, revealing Random Forest as the frontrunner in performance, albeit with its challenges. Chief among these challenges was the prevalent issue of class imbalance within the dataset. To address this critical concern, a diverse array of optimization algorithms, such as Gray Wolf, BAT, and Firefly, were meticulously examined. Moreover, a novel hybrid optimization algorithm, PSO+Gray Wolf, was developed to tackle the class imbalance intricacies inherent in the KDD dataset. The integration of this hybrid approach with Random Forest for classification yielded notably promising outcomes. The proposed model is implemented in Python and results are analyzed in terms of accuracy, precision, and recall.