Projection pursuit fuzzy rules classification evaluation method of commercial credit

摘要

Comparing with the traditional method of credit evaluation, this paper presents a classification evaluation method based on the projection pursuit and the fuzzy rules. Firstly, we use projection pursuit technology to reducing the dimensionality of the training sample, and use genetic algorithm to optimize the projection direction to find the best projection value, classification in accordance with the projection value. Then according to the classification results and the optimal projection value, three types of fuzzy membership functions are obtained by using fuzzy trapezoidal distribution method. Finally, based on the distribution function of testing sample and the fuzzy rules membership function, we calculate the fuzzy nearness degree, and according to fuzzy nearness to determine the credit level of the sample. Numerical examples show that the accuracy of this method is markedly improved compare with the traditional credit evolution method.