基于粗糙集理论的病理诊断规则提取算法研究

Research on pathology diagnosis rule extraction algorithm based on rough sets theory

针对病理诊断规则获取问题,采用基于粗糙集理论的规则提取方法.首先进行连续属性的离散化,用遗传算法对CAIM(class-attribute interdependence maximum)离散化算法进行改进.然后利用粗糙集理论进行规则提取.采用以核为基础的增量式约简算法,综合考虑属性对约简的增益和属性在剩余属性集中的重要性,给出了衡量属性重要性程度的一个准则.随后进行属性值约简,获取诊断规则.

In view of pathology diagnosis rule extraction, we use the rule extraction algorithm based on rough sets theory. Firstly, the discretization of continuous attributes is carried out and the genetic algorithm is used to improve the CAIM （ class - attribute interdependence maximum） discretization algorithm. Secondly, rough sets theory is applied for the attribute reduction. We use heuristic algorithm for attribute reduction based on the cores. Meanwhile, we propose a criterion for measuring attributes importance, which considers both the gain for the reduction and the importance in the remaining attributes. Finally, the value reduction is carried to get the diagnosis rules.