| With the further development of China's economy and finance,the competition among enterprises is more fierce and diversified.However,due to information asymmetry,it is easy to generate enterprise credit risk,which is not conducive to creating a good market environment for fair competition.In recent years,with the rapid development of digital technology and information economy,Internet + and digital technology have promoted the transformation and upgrading of traditional industries at a high speed.Meanwhile,the rapid growth of electronic users has also brought about the explosion of information.Under this situation,our government has promoted the public credit information sharing mechanism,through a unified code of social credit system,and standardized credit information collection,integration,exchange,distribution and use of the method,achieving the connectivity of the local credit information platform,industry information system,the market credit investigation,credit rating agencies with the central database,as well as speeding up the big data resources integration,and defining the boundaries clear in between information privacy and information disclosure.Among them,it is relatively easy to involve private personal data,the enterprise data has the attribute of relatively open and transparent.The country has also issued a series of relevant laws and regulations,which have been standardized and improved,providing a basis for the collection,evaluation and analysis of enterprise big data.All of above laid a solid foundation and a clear direction for enterprise credit risk evaluation and management,however,we should also see that regarding to the current various business scenarios of constant innovation,credit reporting services and related products are still relatively simple,and the prevention and control of enterprise credit risk is also slightly rough,with mainly two ways including credit ratings and enterprise credit.Enterprise rating is mainly targeted at large and medium-sized enterprises,which is a product of qualitative and quantitative analysis based on financial information through on-site research.While Enterprise credit products are mainly issued by the credit information center of the people's bank of China based on the credit information of enterprises credit report,there is no mature domestic enterprise credit score products.Limited by the traditional credit model,most small and micro enterprises cannot be covered by the traditional credit system because they have no credit records in the bank.At the people's bank of China's teleconference for the second half of the year held in August 2019,governor Yi Gang clearly proposed to "vigorously promote the exploration of the use of non-credit data to develop credit investigation services for small and micro businesses and for agriculture,rural areas and farmers,so as to improve the effective supply of credit investigation.” The purpose of this study is to use data mining technology to collect,analyze and evaluate non-financial credit information published on the Internet by enterprises,and try to establish an early warning model of enterprise risk based on non-financial credit information,so as to predict the possibility of abnormal operation risk of enterprises.Through the collection of non-financial credit information,more diversified credit investigation service products are created,and the application scenarios of the existing credit investigation system are further expanded,so as to more comprehensively reflect the credit status of information subjects and promote the completion of social credit system.In view of the above subjects,this paper mainly expounds and discusses the following two aspects:First,to classify the existing enterprise credit information.As non-financial credit information of enterprises is widely distributed in government departments,public institutions and some enterprises.Such information has the characteristics of wide range,low concentration and multi-data dimensions,and there is no unified standard of data interface.This paper firstly sorts out the current non-financial credit information data sources of enterprises and classifies non-financial credit data.According to the complexity and variability characteristics of this kind of information,through the seize of the essence attribute of information,to classify the enterprise credit information through normative and systematic classification method and considering the premise of innovative,to divide the information into 5 big categories including enterprise basic information,enterprise associated information,credit transaction information,credit evaluation and credit auxiliary information.Secondly,to setup the model through testified data and validate data collection,variable design,data cleaning,model building and verification.The data of this paper is coming from the database of a credit investigation and filing institution of a domestic enterprise.According to the method of stratified sampling,and referring to the proportion of the sample size of each province and city from 2018 statistical yearbook "number of enterprise legal person units by region registration and registration type" to obtain the proportion of the sample size of each province.The information of 50,000 enterprises was extracted from the database,and the data from 27 dimensions including industry and commerce,justice,tax and credit China were extracted through the enterprise information classification principle,and 282 characteristic variables were designed.After the analysis of the data,the standards for the bad samples,the observation period and the performance period of the data were formulated,and after the data was cleaned,43 characteristic variables were re-selected.The data was subjected to WOE binning,and the variable IV was calculated.The value calculation and ROC curve verify the discrimination ability,prediction ability and stability of the model,which proves that the model is effective and has a certain prediction ability.This model proves that by modeling the company's non-financial credit information,it is possible to predict the possibility of the company's future revocation.In this paper,a credit score card is designed for the 12 input variables,and an enterprise credit score based on non-financial credit information is issued.In view of the present enterprise credit information sharing model and risk assessment methods,the innovation of this paper mainly embodied in two aspects: firstly,through the national credit information classification standards and the relevant research literature,further carding and summarizing the enterprise credit information classification.This information basically covers the current mainstream of non-financial information classification information data sources,all corresponding category can be found through data field mapping.Secondly,compared with the previous enterprise risk warning model,all the enterprise information in this paper is from the Internet public information,and all of them are non-financial enterprise basic information,enterprise correlation information,credit auxiliary information and credit evaluation information.With the construction of the social credit system and the development of the Internet,a large number of public information,including non-financial credit information of enterprises,will be publicized.Based on this information,more enterprise credit investigation products can also be developed and serve as an important supplement to credit information investigation products.It can not only achieve the effective supply of credit information,but also to predict the risk level of enterprises comprehensively,and it is an active attempt to explore the use of non-credit data to establish an early warning model of credit risk for providing the credit information service. |
PDF Full Text Request