Research On Enterprise Portraits For Credit Reporting

The rapid development of market economy has brought new opportunities and challenges to modern enterprises.It can not be ignored that the credit information system of enterprises in our country has not been perfected.In particular,the problem of closed credit information and weak anti-risk ability of SMEs is particularly prominent and needs to be resolved.In order to speed up solving the problems such as incomplete credit information of enterprises and unclear cognition of economic subjects to enterprises,the improvement of the enterprise credit reporting system has become a crucial link in China's current economic development.Aiming at these problems existing in the field of credit information at present,by deeply studying the data characteristics under the background of credit information and the related theories of enterprise portrait,this paper applies the technology of enterprise portrait to the field of credit information,and constructs the system frame of enterprise portrait.On the basis of this framework,this paper further studies the method of extracting the rating label of enterprises in the credit field.The main work of this paper includes the following aspects:Firstly,The primary task of enterprise portrait is to clarify the purpose of portrait and the characteristics of enterprise data.Combined with the actual situation of the enterprise portraits applied to the credit reporting scenarios,a set of framework of enterprise portrait system for credit reporting has been designed.The system framework includes the overall framework of enterprise portrait,the label system of enterprise portrait,the credit evaluation index system and the generation method of enterprise portrait label.The framework of the system is the overall implementation process of the enterprise portrait in this paper,and is the key to the portrait.Secondly,Due to the importance of solvency in enterprise credit,this paper studies theextraction of solvency rating label based on XGBoost.Firstly,the enterprise credit evaluation index data is obtained through the network crawling method,and the weight analysis based on the entropy method is performed.Based on the weight analysis results,the importance of the solvency of enterprises in the credit reporting system is determined.Then,the entropy method is used to obtain the solvency score of the enterprise,and the unlabeled sample data is labeled.Finally,the research implements the XGBoost-based solvency rating model to extract the solvency rating label of the enterprise.Several experiments are designed and compared to verify the effectiveness and accuracy of the model.Thirdly,a method of improving Light GBM by using Bayesian hyper-parameter optimization algorithm is proposed.In this paper,the improved Light GBM model is applied to extract the credit risk rating tags of enterprises.The method uses Bayesian optimization algorithm to update the historical records through iteration,and finally finds a set of optimal hyper-parameter settings for Light GBM.This paper compares experiments from different angles,and evaluates the performance of the proposed model by using several evaluation methods.The experimental results show that the model has higher accuracy,and its classification performance is superior not only to the traditional rating classification model,but also to the Light GBM improved by other hyper-parameter algorithms.