基于聚类分析的决策表约简

基于粗糙集方法提出了一种系统的决策表约简和决策规则提取方法. 为了避免现有属性离散化方法的不足,使用多元统计中的聚类分析,并借助树形图,R2、半偏相关以及伪F统计量,对连续属性进行离散化处理,得到适合粗糙集方法要求的决策表. 在此基础上,简化了基于可辨识矩阵和逻辑运算的传统属性约简算法,并完善了启发式算法进行属性值约简和决策规则提取. 最后,以应用实例验证了该方法的可行性和有效性.

Decision Table Reduction Based on Cluster Analysis

Based on the rough set theory,a new systematic method is proposed to reduce the decision table and induce the decision-making rules.In order to avoid the shortcomings of current discretization methods,the cluster analysis of multi-variable statistics is introduced to discretize the continuous attributes in the decision table.With the dendrogram and three useful statistics,i.e.R~2,SPRSQ and PSF,the decision table is derived which can meet the requirement of the rough set theory.After that,the traditional algorithm of attributes reduction based on the discernibility matrix and logical operation is simplified,and an improving heuristic algorithm for attribute value reduction and decision-making rule induction is presented.Finally,an illustrative example is proposed to validate its feasibility and(effectiveness.)