置信规则库结构识别的置信K均值聚类算法

A Belief K-means Clustering Algorithm for Structure Identification of Belief-rule-base

针对置信规则推理作为系统控制器时的应用,提出一种置信K均值聚类算法用于置信规则库的结构识别。在构建好置信规则库的推理框架后,该算法通过对规则前项输入变量的历史数据进行挖掘,得到合理的置信规则库结构,提高推理与决策的精度。相对于传统专家知识确定置信规则库结构的方法,该算法的特点是:最优聚类与相邻评价等级之间的距离成正比,与人的认知能力相一致;最优聚类保证采样点以最小的距离靠近评价等级,也就是保证输入变量尽可能趋近置信规则前项。通过置信规则推理在集约生产计划中应用的案例分析验证了该算法的合理性和有效性。

A belief K-means clustering algorithm is proposed to identify the structure of a belief-rule-base for belief-rule based reasoning in system control.After the inference framework of the belief-rule-base is constructed,the algorithm can generate a reasonable structure of the belief-rule base and improve inference accuracy and decision quality through mining historical data about antecedent input variables.Compared with traditional expert-knowledge based methods for determining the structure of belief-rule-base,the new algorithm has the following characteristics.The generated optimal cluster is directly proportional to the distance between two adjacent evaluation grades,which is consistent with human cognition.The optimal cluster ensures that the sampling data are around the evaluation grades with minimum distances,which ensures that input variables optimally approximate the antecedents of belief rules.A case study is conducted to apply the belief-rule based reasoning to aggregate production planning,which demonstrates the rationality and effectiveness of the proposed algorithm.