基于时间序列的模糊聚类与规则提取信用评价模型  

Credit Evaluation Model Based on Time Series Fuzzy Clustering and Rule Extraction

作　　者：张洪祥[1] 毛志忠[1,2] 

机构地区：[1]东北大学信息科学与工程学院,辽宁沈阳110004 [2]东北大学流程工业综合自动化重点实验室,辽宁沈阳110004

出　　处：《东北大学学报（自然科学版）》2010年第4期465-468,共4页Journal of Northeastern University(Natural Science)

基　　金：国家自然科学基金资助项目(60774068)

摘　　要：提出基于多维时间序列模糊聚类与模糊规则提取技术相结合的模糊分类系统,将其应用于信用评价研究.该方法利用投影寻踪技术对多维时间序列数据进行降维处理并进行模糊分类;根据分类结果和最佳投影值提取模糊规则,采用梯形分布法生成三个模糊隶属函数;最后根据计算模糊贴近度确定样本的信用级别.实例证明该方法具有良好的评价效果和实用价值.A fuzzy classification system was presented by combining the multidimensional time series fuzzy clustering with rule extraction to evaluate the credit.This method was applied to the dimensionality reduction of multidimensional time series via projection pursuit technique for fuzzy clustering.According to the results of the classification and the extraction of fuzzy rules through best value projection,three types of fuzzy membership function were generated using trapezoid distribution method.Then the sample＇s credit level was determined according to the fuzzy approximability in computation.Exemplification results showed that this method is good at evaluation effectiveness and practicality.

关 键 词：多维时间序列 投影寻踪 模糊聚类 模糊规则提取 信用评价 

分 类 号：F830[经济管理—金融学]