基于多维时间序列的灰色模糊信用评价研究

Research of multidimensional time series credit evaluation based on gray-fuzz analysis model

传统信用评价技术多在孤立时间点上对受评目标数据进行评价分析,但受评目标由于某些原因可能产生数据"突变",导致评价结果失真,产生信用风险.针对这一问题,本文提出应用多维时间序列数据对受评样本进行信用评价.该方法首先对多维时间序列数据使用灰色关联分析方法进行分割处理,解决"维数灾难"带来的严重影响,并将得到的灰色关联度值作为信用评判值;再运用模糊聚类方法对信用评判值构成时间序列矩阵进行信用评价分析,得到受评样本的"真实"信用等级.通过实例验证,该方法可以从时间序列的角度观察受评目标信用等级的状态趋势及"波动"情况,解决因为数据"突变"造成的评价结果失真问题,具有良好的评价效果和实用价值.

Traditional credit evaluation technique only evaluates and analyzes the objective data at the isolated time points, but the objective may generate data variation due to some reasons. Therefore, this may lead to infidelity of the evaluation results and serious credit risk. Aimed at this problem, the multidimensional time series data is used to credit evaluation in this paper. First, the multidimensional time series data is dealt with separately by grey relational analysis, in order to solve the serious effects caused by dimension disaster. At the same time, the credit evaluation value is obtained. Then the time series matrix is used to fuzzy clustering analysis . The real credit evaluation grade is got. The results of our experiment showed this method can observe the credit grade fluctuation of the credit evaluated in view of the time series. Thus, the infidelity of the evaluation results caused by data variation problem is solved, and the proposed method has accurate evaluation effect and better performance in practical application.