The Research On Credit Evaluation Of Small And Micro Enterprises Under Incomplete Information

Small and micro enterprises(SMEs)are the driving force of economic and social development and the important force of market players in China.They play an important role in national economic growth,technological innovation,employment problems and social stability.Good development of SMEs is one of the government's targets in our country,which can't depart from the support of financial institutio ns(FIs).At present,financial institutions generally use financial data for credit evaluation to implement lending and risk control,while for SMEs,there is generally little or no such data,that is,the "incomplete credit information" problem.Even in big data era,with other data sources introduced,this problem is more prominent.In this paper,the basic concepts and theories of information,credit risk and data loss are briefly introduced.Then,starting with the characteristics of small and micro enterprises and the causes of credit risks,the lending demand is characterized by ?shorter-term,lower amount,high frequency and stronger timeliness?.This paper analyzes the reasons for the incomplete credit information of small and micro enterprises in view of the different operation and risk characteristics.Finally,combined with the practice of credit evaluation,this paper studies the following three types of typical credit evaluation problems under incomplete information scenarios:Firstly,there is no historical data in the early stage of credit business development,so we cannot use statistics and data mining to build the model of incomplete information.Considering the shortcomings of the traditional expert empirical method in collecting,processing and synthesizing expert opinions,this paper studies the credit evaluation model construction method based on group decision making—3-SEIJ.3 scale method is used to build the discriminant matrix to determine the index weight,and 3 to 9 scale mapping tec hnology can decrease the difficulty of expert judgment.Also,the consistency adjustment technique is applied to solve the problem of repeated judgment caused by the inconsistency of judgment matrix.Build problems of credit evaluation model,which is lack of historical information in the early stage of credit business,can be solved by expert and trusted weighted clustering technology.Secondly,although there is some accumulation of default data during the development of credit business,the information i s still insufficient to improve the incomplete information of the evaluation model.In this paper,AHP and neural network are combined to upgrade the model online,which is based on a small amount of default data,so as to make full use of the continuously complete default data and experts' empirical knowledge information,and solve the problem of the randomness of the initial assignment of neural network model,the comprehensibility of neural network and premature convergence to the local solution.The eff ectiveness of the method in improving the performance of the credit evaluation model is verified by experiments.Thirdly,due to the incomplete situation of missing information items in the assessed object,this paper studies a new data filling algorithm b ased on the iterative idea of EM algorithm in the specific credit evaluation process after the establishment of the evaluation model.Two key stages of the algorithm,namely the construction stage of the univariate predictive model library and the iterativ e filling stage,are described.This algorithm solves the problem of multiple variables missing at the same time,and abandons the dependence of EM algorithm on probability distribution hypothesis,so it has wider applicability.Then,the method is compare d and analyzed with the general mode fill,mean fill and EM fill through the fill-reduction experiment and post-fill credit prediction experiment of three benchmark credit assessment data sets.Finally,this paper analyzes and discusses some other relevant measures and methods to deal with the situation of incomplete information in the process of credit evaluation of SMEs,and proposes the concept of an evaluation model assisted development system.Research ideas and results of this article could promote th e development of credit evaluation technology,and improve financial institutions credit evaluation work efficiency,optimizing the resources distribution of financial institutions.This will enable financial enterprises that can truly create social wealth to obtain financial support,thus providing effective decision-making basis and technical guidance for the country's economic transformation and upgrading and the implementation of national innovation strategies.