Spam Email Detection Using HAN-LSTM Network Optimized with Novel Hybrid Bonobo Optimizer and Hunger Games Search Algorithm
Indu B Singh, Hemanth Siddharth Desugari, Shaurya Wadehra, Venkata Sai Karthik · 2024
In today's digital world, spam email has grown to be a serious problem that causes problems for email service providers as well as subscribers. Creating a reliable system that can effectively filter out spam emails is required to address this problem. This paper proposes a novel approach to spam email detection that combines advanced feature engineering and hybrid metaheuristic optimization. The proposed method employs Bidirectional Encoder Representations from Transformers (BERT) to derive dense feature vectors, followed by a HAN-LSTM network for improved feature representation and classification. Additionally, we introduce a novel hybrid metaheuristic optimization algorithm comprising of Bonobo Optimizer and Hunger Games Search (hBOHGS) that combines the collaborative exploration of Bonobo Optimizer with the competitive selection of Hunger Games Search to efficiently find optimal values for the parameters in the HAN-LSTM network. Our experimental evaluation on the Enron 1 spam dataset demonstrates the effectiveness of our methodology, achieving an accuracy of 98.04%, a recall of 0.99, and a precision of 0.98. These findings highlight the practical applicability of our system in effectively detecting spam emails, providing a promising alternative in the field of spam email detection.