A Comparative Study of Different Bayesian Approaches to Anti-Spam Filtering

摘要

This paper has dissected the concrete effects and related details of different Bayesian approaches in email classification. First, this paper put forward the email classification model and the basic assumptions. Then we discussed the email features extraction and selection issues, especially focused on the documents frequency and information gain methods, and simultaneously performed the features reduction to the rule of thumb. Finally, a comparative study has been driven on three different kinds of algorithms based on Bayesian methods, and so the influences caused by different feature selection methods.