Three-way Email Spam Filtering with Game-theoretic Rough Sets
Zhang Yan, Pengfei Liu, JingTao Yao · 2019 International Conference on Computing, Networking and Communications (ICNC) · 2019
-Three-way email spam filtering reduces the chance of misclassification by dividing incoming emails into three folders, i.e., a spam folder consisting of email that we believe are spam, a legitimate folder (inbox) consisting of email that we believe are legitimate, and a suspicious folder consisting of email that we cannot make sure decisions based on available information. Game-theoretic rough sets (GTRS) are a recent proposed model for determining three-way classifications from a tradeoff perspective when multiple criteria are involved. This paper applies GTRS in email spam filtering to obtain the suitable three-way email spam filtering which represents a tradeoff between the accuracy and coverage. Game formulation and repetition learning mechanism of GTRS are examined in detail. The experiment is conducted based on UCI Spambase Data Set. The three-way email spam filtering determined by Pawlak rough sets and GTRS are compared. The experimental result shows that the three-way filters determined by GTRS are able to improve the coverage level significantly by sacrificing the accuracy level a bit.