Based On The Logistic And Personal Credit Assessment Of The Neural Network Combination Model Research

In the era of rapid development of market economy, personal credit, as the necessary precondition of the individual excessive consumption, plays a vital role in expanding domestic demand and stimulating the development of financial and credit industry. In the other hand, personal credit as a carrier of personal act has been gradually recognized as the necessary means of upgrading personal morality, and maintaining economic order and social order. So the importance of personal credit assessment is evident. However, in the West countries, personal credit rating system has entered the stage of a specialization and industrialization, the construction of personal credit assessment system has only just started in China. In the process of individual consumer credit development, one of the major problems is that individual credit risk is difficult to assess and control. In recent decades, all commercial banks take consumer credit business as an important part of its development strategy, which makes the study of individual credit assessment methods become particularly critical.In view of that most current personal credit assessment models only take into account the model accuracy, from a systematic perspective, four aspects would be taken into account--accuracy, robustness, time complexity and interpret ability, and the personal credit assessment quantitative model would be focused on. To begin with, an evaluation index system is built and a preliminary analysis with the sample data is conducted. After data pre-processing, the logistic regression model is constructed and a deeper explanation of the statistically significant individual credit evaluation indicators is conducted with the regression result. Aiming to the problems of BP neural network like time-consuming, we use the Boosting algorithm and an improved LMBP algorithm to optimize it. After the optimization, there is an improvement in the classification accuracy on the personal credit data. Through an applicability analysis to these single models, a Logistic-ILMBP combination model based on entropy method is proposed. Last but not the least, the results are compared and analyzed.Research indicates that both two improved approaches to BP neural network can effectively improve the classification accuracy, but the improved LMBP (ILMBP) algorithm has more advantages in performance and efficiency, therefore, more practical value. The logistic-ILMBP combination model based on entropy method combined the high classification accuracy of neural network model and the robustness of the logistic regression model. As the principle of information entropy is used in weight calculation, it is easy to explain. In the early stage of development in quantitative model of personal credit assessment, our country is also facing the problems like incomplete data. Logistic-ILMBP combination model can not only achieve higher classification accuracy, but also has good reliability, so it has a practical significance for both personal credit rating agencies and decision-making institutions.