SME Credit Rating Model Based On Integrated Learning

Risk is a measure of the likelihood of future losses.In social life, risk is everywhere.Risk will bring political change or changes in the weather, which challenges different interestgroups. Risk prediction is an important topic in financial activities. SME（Small andmedium-sized enterprises） credit rating is a research of the ability of SME repaying the loanin the future.In other words, it is a predict of the default risk.With the development of economy and society, the financial markets have becomeincreasingly active. More and more experts’ and scholars’ attention have been attracted by theresearch of risk management in financial activities. They come from different fields, such ascomputer science, finance, economics and so on. SME as an important force in the financialmarkets is becoming more and more important in the development of national economy. It isindispensable that SME play in the role of promoting industrial transformation and upgrading,innovating technology and improving international competitiveness. However, due to thecharacteristics of China’s SME and the lack of experience in risk management as well as thelow level of China’s financial industry development, it is difficult to evaluate true credit statusof SME. In order to avoid the risk，financial institutions are reluctant to loan to SME, whichresult in a severe problem of financing and hampered the development of SMEs’ growth. Thekey to solve this problem is to establish a specific SME’s credit rating model. Research onthis topic in the country has just started, so it has a higher value of application and broadresearch space.The author read a lot of literature about corporate credit ratings and he have fully studiedclassification model prediction in the field of machine learning as well as the knowledge offinancial aspects on risk management and risk identification. Based on these, the authorconducted in-depth research on selecting the index factors and building the credit ratingmodel.First, this paper introduced the status of research of SME credit rating. Then it analyzedand summarized the current index factors affecting SME credit. Later the author comparedand analyzed classification learning algorithms and models widely used in credit ratings. Theauthor discovered the advantages and disadvantages of these methods. On the basis ofprevious relevant results, according to the low prediction accuracy rate and the insufficientgeneralization ability of models in the current credit rating system the author has done a lot ofresearch in terms of SME credit rating. And he put forward a model framework of this article.The main content of this paper is composed of the following two aspects:(1) Because the research on SME credit rating is very few. The author referred to a number of factors affecting the SME credit when building the index system. Then accordingto the characteristics of SME, the author summarized the factors which affect credit status ofSME. And he adds them to the index system of SME. Based on previous research, the authorbuilt a set of index system affecting SME’s credit.(2)Inspired by the idea of ensemble learning, the author abandoned the previous creditrating model. He combined several homogeneous models in the field of machine learninginnovatively. The paper built an evaluation model specifically for SME’s credit rating. Onthe basis of SME credit evaluation index system proposed by the author, the author studiedthe system of SME credit evaluation model based on integrated learning systematically. Thebasic learning machine selected by the Integrated-Learning model contains Decision Treealgorithm、BP-neural network algorithm、SVM algorithm、Linear Regression as well as Na veBayesian. He tests the effectiveness of the model in the pilot phase.