对模糊聚类分析法的改进及其在SRM中的应用

针对传统模糊聚类分析法在信息系统的决策分析中无法有效解决各因素之间的相关性干扰,以及不同特征属性对聚类目标存在重要性差异等问题,本文提出一种融合层次分析法、Mahalanobis距离法及专家群决策法的改进模糊聚类分析法。在特征属性的重要性处理环节,层次分析法用于判断不同特征属性的相对重要性差异;引入Mahalanobis距离法进行相似矩阵的构建,能解决变量之间的相关性干扰问题;专家群决策法用于确定最佳阈值λ,能最大程度地降低主观因素对评价结论的不利影响。在SRM中的应用实验结果表明,改进的模糊聚类分析法在客观性和准确性上更能满足信息系统决策分析的需要。

Modification of the Fuzzy Clustering Analysis Method and Its Application in SRM

In the decision analysis of information systems,the traditional fuzzy clustering analysis has some shortcomings,which not only can not resolve the relevance interfering problem of the characteristic attributes,but also ignores the difference of importance dimensions among the different characteristic attributes.In order to overcome these defects,an improved algorithm is proposed based on the analytic hierarchy process,the Mahalanobis distance algorithm and the group decision method.In the link of importance treating,an analytic hierarchy process is used to estimate the importance dimensions of different characteristic attributes.The Mahalanobis distance algorithm is introduced to build the similarity matrix,and it can resolve the problem of relevance interference of the variables.The experts group decision method is also introduced to decide the best threshold λ,through this method,the adverse effects of the subjective factors to the conclusion of the analysis have been reduced greatly.The application in the supplier relationship management system shows that the improved fuzzy clustering analysis is greatly satisfied with the needs of the decision analysis in information systems.