Research Of Credit Risk Based On Information Fusion

At present, importance of personal consumer credit is essential to development of economics. As an inevitable result of socio-economic development, credit has become an integral part of modern society operation. A good credit can not only sustain economic relationship, but also protect the whole order. Therefore, research of the personal credit system and evaluation algorithm is becoming more significant. However, the relevant assessing system has not caught up. This paper studies credit risk problem by using information fusion, its main work is as follows:First of all, common theories and methods of credit risk evaluation are simply introduced. Researching background is shown as well.Secondly, principles and applications of Support Vector Machine based on SRM and K Nearest Neighbors is presented. Then methods and theory of multiple classifiers fusion. is introduced.At last, SVM and KNN are combined based on the classifiers fusion to form the SVM-KNN algorithm. Effectively, this algorithm can solve the overfitting problem and high error scoring rate caused by overlapping samples besides the optimal interface. Meanwhile, in order to eliminate redundancy in the dataset, data-preprocessing is taken according to the information provided by distance between positive and negative. Precise steps are given and an assessment model is built on it. In empirical analysis, after German Credit Database is preprocessed, an experiment is carried out and the result indicates that SVM-KNN model has better classification performance and efficiency than SVM itself.