Research On Small Business Credit Scoring Based On Ensemble Decision Tree Algorithm

Small businesses are an important part of China`s economy, and also anindispensable force in the socialist economic construction. Since reform and opening up,small businesses has increased rapidly with the establishment of market economicsystem，and small business has played an important role in the optimization of industrialstructure, increasing employment, technological innovation, promoting social harmonyand stability. However, small business still faces lots of problems, such as high cost,immature market, low technical content, especially the financing problem. At present,China’s small business mainly rely on bank financing, but the traditional credit scoringsystem of commercial bank is established for large enterprises, and it is not suitable forsmall business which is with low credit data validity, complex factors and imperfectfinancial documents. Therefore, the establishment of a low risk, high precision, highefficiency personal credit scoring model for small business can not only solve thefinancing problem of small business, but also can improve profitability and promote thedevelopment of credit business of commercial banks.Based on the study of credit scoring method and application, this paper analyzessmall business deeply, and points out that the credit scoring on small business has tosolve the problems of low credit data validity and complex factors. Therefore, DecisionTree is used in credit scoring of small business, and the advantages and limitations ofthe model are proposed after the empirical test. To solve the problem of redundancysamples, Particle Swarm Optimization algorithm is used to get the reduction ofattributes. To solve the problem of not determinable, the ensemble Decision Treealgorithm based on Bagging and Random Subspace is used to improve the capability ofidentifying credit risk. The empirical test shows that PSO-ensemble Decision Treealgorithm can keep the advantages while solve the problems of Decision Tree when it isused in the credit scoring of small business, and the accuracy and stability of newalgorithm is also improved.