This paper explores the design of effective spam filters using combined Näive Bayes classifiers. We discuss various tokenization methods for extracting valuable features from emails, creating diverse training sets for individual Bayesian classifiers. Through computer experiments on our spam dataset, we compare different fusion methods based on class labels and supports to determine the best approach for ensemble evaluation and establish our final proposition.

Project link: https://link.springer.com/chapter/10.1007/978-3-642-33018-6_26