부산대학교 도서관

As microfinance companies that has focused on improving rural finance and helping SMEs to solve financing difficulties, it has received great attention from governments, scholars, and related institutions since its emergence. Therefore, the study of microfinance companies is of great research significance both in theory and in reality. However, because the object of payment is a special group and the loan target is operating an inefficient and vulnerable industry. There are many factors constrained by the outside world, leading to frequent occurrence of various types of debt escaping and high loan default rate. All these have caused a higher risk of micro loan business. Therefore, how to establish a credit risk assessment model to reduce the credit risk of microfinance becomes an urgent issue at present.At present, there are too many human factors in the customer credit risk rating method adopted by China's financial industry, which affects the accuracy and scientificity of the rating system. The classification and71forecasting technology using data mining can analyze the organization's original data highly automatically. Data mining models can make inductive reasoning and automatically classify customers' credit risk levels. Data mining technology is an emerging discipline that obtains useful knowledge from large amounts of data. Introducing data mining technology into credit assessment models can solve some of the current credit risk issues faced by the microfinance industry. Building a personal credit assessment model can effectively help microfinance companies improve personal credit efficiency and accuracy, and improve personal credit product quality and risk management capabilities. This research compares several commonly used risk assessment models and proposes a personal credit assessment model based on data mining technology. The use of data binning method to filter the data attributes, to find out the attributes that have a greater impact on the classification forecast modeling. Decision models, neural networks, and logistic regression methods were used for analysis modeling. An empirical analysis of the model using personal consumer loan data from a small loan company in China shows that the model has a low degree of deviation and goodpredictiveresultswhentheoriginalinformationisnotsufficient.