Abstract: This paper proposes a multistage three-way decisions of spam SMS (short messaging service) filtering model. By using different granularity information to express the SMS, with the sequence decisions (i.e., multistage, multiple steps in decision-making), at different decision-making stages, based on different information granularity respectively three-way decision-making, those can avoid to make irrational decisions when the message size is too big or insufficient information. If the information does not meet the current decision-making conditions, this model can add enough granularity information before making a further decision. Finally, the experimental results prove the rationality and validity of the model.

Key words: three-way decisions, sequence decision, short messaging service (SMS), multistage spam SMS filtering, granular computing

摘要： 提出了一种多阶段三支决策垃圾短信过滤模型。该模型使用不同的信息粒度对短信进行表示，运用序列决策（即多阶段、多步骤决策），在不同的决策阶段基于不同的信息粒度分别进行三支决策，有效地避免了当信息粒度太大或信息量不足时进行不合理的决策，对于不能满足当前决策条件的信息，可以通过补充足够的粒度信息作进一步的决策。最后通过实验证明了该模型的合理性和有效性。

关键词: 三支决策, 序列决策, 短信（SMS）, 多阶段垃圾短信过滤, 粒计算