| Industrial credit evaluation is of importance in building Industry Credit System of China. Since the construction of the system has just begun, there is no standard to follow. And the researches about credit evaluation in foreign countries are mainly restricted in traditional credit risk assessment, which does not concern about characteristics of the industries in China, so that they cannot be directly applied. Crain and oil industry, as one of the pilot industries in building Food Safety System of China, has abundant industrial characteristics. To take grain and oil industry as a foothold to start researching on industrial credit evaluation model can provide a way for other industries.After analyzing the characteristics of grain and oil industrial credit evaluation, and comparing the applicability of existed evaluation model, a new evaluating course, circulating evaluating course, is proposed and emphasized in this paper. Based on this course, a model combining experts mark method and recurrent neural network is developed. After specifying the features of each phase in the course, several technologies are effectively utilized. Adaptive learning with momentum algorithm and Bayesian Regularization method are used in Score Evaluation Phase and evaluation optimization phase respectively, in order to increase the converging speed of network training and promote the generalization of the network. Fuzzy mathematical theory is applied in detailed Grade Division Phase, in order to reduce the subjective judgment of the experts, and make the results of the evaluation easy to describe and publish. Principle Component Analysis is utilized in Measures Decision-making Phase, in order to reduce interference from the data with no contribution. Regression Analysis is used in Result Analysis Phase to give a measure of whether to continue the loop or jump out of it.Through combining of experts mark method and recurrent neural network, this model make the network learn the evaluation knowledge of the experts, and reduce the subjectivity of evaluation from human judge, so that the conflict between vacancy of the evaluation standard and overabundance of human interference is relieved.Based on this model, application research is carried out on Grain and Oil Industry Credit Evaluation Platform, and the directions on optimization of the theoretical model and engineering application are pointed out in the conclusion. |
