A Fusion-Based Approach with Bayes and DeBERTa for Efficient and Robust Spam Detection
Ao Zhang, Kelei Li, Haihua Wang · Algorithms · 2025
Spam emails pose ongoing risks to digital security, including data breaches, privacy violations, and financial losses. Addressing the limitations of traditional detection systems in terms of accuracy, adaptability, and resilience remains a significant challenge. In this paper, we propose a hybrid spam detection framework that integrates a classical multinomial naive Bayes classifier with a pre-trained large language model, DeBERTa. The framework employs a weighted probability fusion strategy to combine the strengths of both models—lexical pattern recognition and deep semantic understanding—into a unified decision process. We evaluate the proposed method on a widely used spam dataset. Experimental results demonstrate that the hybrid model achieves superior performance in terms of accuracy and robustness when compared with other classifiers. The findings support the effectiveness of hybrid modeling in advancing spam detection techniques.