A Cost-Sensitive Enhanced Whale Optimization for Network Traffic Classification Using Fully Connected Encoders

Network traffic, as an important element of modern network, requires proper classification to ensure security, performance, and compliance. Analysis of service and capacity planning lies on precise classification for business evaluation and strategic planning. Therefore, the proposed work explores the process of classification that can classify the applications. Classification and feature selection are extremely crucial in improving the models performance in different domains. The proposed work presents a new approach that integrates pre-processing, Enhanced Whale Optimization Algorithm (E-WOA) for the purpose of feature extraction and Fully Connected Encoders (FCEs) for classification. The EWOA effectively chooses the most informative features, dimensionally reducing the essential information without any loss. These features are then subjected to an FCE-based classifier to provide high accuracy and robust decision-making. Experimental results confirm the outperformance of this hybrid model over traditional feature selection and classification techniques. The method proposed in the given work always attains excellent performance of 0.981 and 0.982 for VPN and non-VPN datasets respectively. Which gives a faster and scalable solution to high-dimensional data classification and feature extraction. It presents a prospective advantage for many applications in real-world situations and for varied disciplines such as finance, bioinformatics, and healthcare. Ideas for future research might include examination of more optimization methods and combination with other metaheuristic methods