Research On Enterprise Credit Assessment Based On Model Fusion

With the rapid development of Chinese socialist market economy,credit risk evaluation is more and more important to simulate the vitality of market entities,to promote highquality market development,to prevent credit risks,and to maintain financial order.As an important part of the market economy,enterprises play an important role in the development of the socialist market economy,therefore,the research on enterprises credit evaluation is of great significance.At the same time,with the development of big data technology,applying big data technology to local government decision-making will also have important research significance for improving government administrative capabilities,reducing operating costs,and building a smart government.Based on some enterprises data in Qingdao provided by the Qingdao Municipal Government in the Shandong Province Data Application Innovation Contest,this thesis study the credit status for some enterprises in Qingdao by using modern statistical analysis techniques,machine learning methods and data mining methods for empirical analysis.First,we introduce the significance of the research on enterprises credit evaluation in the development of the current socialist market economy,the status of credit evaluation research at home and abroad,and the development process of credit evaluation research.Then we briefly show the basic ideas and facts of the XGBoost model,the Light GBM model,random forests,the Logit model and the BP neural network,the principle of stacking fusion method,and the model evaluation indicators.Finally,we analyze the sample data set and do the feature processing,basing on the evaluation indicators such as ROC curve and AUC value,we successively construct and analyze the XGBoost model,the Light GBM model,random forests,the Logit model and the BP neural network in turn.In order to further improve the performance of the model,a stacking method is used for model fusion which is compared with the single model.The research results show that although the time-consuming increases after model fusion,there is a certain improvement in AUC,which is suitable for enterprises credit evaluation in the case of imbalanced categories.The research attempts of this thesis provides an idea for the construction of enterprises credit evaluation models.