A decision-theoretic rough set approach to spam filtering

Abstract

Spam filtering is a research hotspot of information security. For the weak fault-tolerant ability of traditional filtering methods, an approach to spam filtering based on α - positive-region of decision-theoretic rough set (DTRS) is developed. Firstly, α -positive-region attribute reduction theorem is adopted to reduce email attributes. Then, according to the minimum risk Bayesian decision theory, a three-way decision, named spam, doubt and non-spam, is realized by depicting the undecided emails using boundary region of DTRS. The simulation results show that this approach is effective and helpful to improve the performance of spam filtering.