Research On MTS Based On K-means Algorithm And Its Application In Personal Credit Evaluation System

With the rapid development of economy, credit transactions become more and more common. To improve the technology level of credit management and establish a good credit evaluation model are very important. Mahalanobis-Taguchi System is an effective pattern recognition method. It has been applied in many fields in recent years. This article introduces k-means clustering algorithm into the Mahalanobis-Taguchi System, and builds a Mahalanobis-Taguchi System based on k-means clustering algorithm. This article also studies its application in the evaluation of personal credit and achieves good results. The main researching content includes two aspects as following:(1) Based on k-means clustering algorithm of Mahalanobis-Taguchi System theory researchFiltering characteristic variables is an important part of Mahalanobis-Taguchi System. Different characteristic variables constituting a benchmark space will get different category analysis. How to effectively filter characteristic variables is a problem which is worthy of studying. Traditional Mahalanobis-Taguchi System using orthogonal table to design test plan, using signal-to-noise ratios to select characteristic variables, thus effectively reducing the number of characteristic variables. This paper attempts to combine k-means algorithm and Mahalanobis-Taguchi System, use orthogonal table to design experiment scheme, for each kind of test plan, use k-means algorithm to cluster original samples, get groups of the accuracy of clustering results, with signal to noise ratio of the accuracy as the evaluation index, and then get the effective variables, achieve the goal of optimization of benchmark space.(2) Based on k-means clustering algorithm of Mahalanobis-Taguchi System application researchThe application research use a bank personal credit evaluation data as the background. The k-means clustering algorithm and improved Mahalanobis-Taguchi System applied in the personal credit evaluation system are carried on the detailed research, comparison and analysis. Finally draw the conclusion:Mahalanobis-Taguchi System based on k-means clustering algorithm model and k-means clustering algorithm by using the method of exhaustion have selected nearly the same characteristic variables, illustrating that characteristic variables selected by Mahalanobis-Taguchi System based on k-means clustering algorithm model are effective; Mahalanobis-Taguchi System based on k-means clustering algorithm model has a higher accuracy in prediction in the sample credit status than traditional Mahalanobis-Taguchi System.