Grouping co-occurrence filtering based on Bayesian filtering

Abstract

Recently, many people are using communication tools on the Web, but some send harmful information to others. Most operators manually deal with harmful information, which is expensive. In this paper, we implement two-word co-occurrence filtering by applying the Bayesian filtering method as a spam filter. We propose grouping co-occurrence filtering based on Bayesian filtering and experimentally verify our approach. Grouping co-occurence filtering detect harmful or safe documents at low cost. Our result suggests that grouping co-occurrence filtering is more stable and has a higher accuracy than co-occurrence filtering baesd on Bayesian filtering. © 2012 Springer-Verlag.