Research Of Small And Mediem Sized Enterprises Credit Risk Evaluation System And Model Based On Neural Network And Rough Set Theroy

Small and medium sized enterprises (SMEs) occupy an increasingly important position in China's national economy. The total number of SME has increased year by year. They also promote some growth such as industrial output, exports, tax and providing jobs for the community. However, the financing problem has plagued the small and medium enterprises. The most of important one is the difficult problem of credit financing.Our government has recently issued policies for continuously guiding and encouraging banking institutions to increase credit support of SMEs. These policies effort to promote the growth of SMEs. While to solve the problem, Commercial banks are also facing the greatest threat comes from the SME credit risk. Credit risk assessment of SME is the critical content for solving the information asymmetry problem between enterprises.From the practice of China's banks, the model of credit risk evaluation is in its infancy. Due to lack of historical data, our commercial banks generally do not establish a quantitative model for credit assessment.This article proposes a SME credit risk evaluation model. It is based on Rough Set and Neural Network. The major disadvantage on the establishment of personal credit rating process is that there are a lot of subjective factors in it. This Model aimed to resolve this disadvantage makes use of the Rough set attribute reduction method in the establishment of credit risk evaluation index system. Then this article will give a more scientific and practical BP neural network model. SME credit rating system is the input of BP neural network model. It uses small and medium-sized enterprises credit history data in the commercial bank credit management information system (MIS) databases for training and learning. This can adjust the weight among neuron in the model and determine the intrinsic link between the input and output.First, the model weakens the human factors between credibility index and weight and improves the accuracy of the assessment results. Second, the model has the powerful non-linear processing and generalization capability. It could be solve the problem which people determine the weight factors and ensure the accuracy and objectivity of the evaluation results.