Analysis And Prediction Of Customer Overdue Based On Bp Neural Network

In recent years,international trade has become more prosperous and some problems have gradually emerged,and the term "subprime mortgage crisis" has again caused much concern.One of the important reasons for the “subprime mortgage crisis” is that various lending institutions,such as traditional commercial banks,and P2 P network credit institutions which develop rapidly in recent years have failed to provide a comprehensive and scientific enough assessment of credit customers.They lack a set of suitable evaluation mechanism and models.Problems such as bad debts and overdue contracts have become more serious.Especially in the 21 st century,the era of big data,“immediate pleasure” has become more and more people's choice,personal credit business has also become more prosperous,the corresponding data dimension has increased greatly,the traditional scoring method can no longer meet the corresponding demand.The essence of credit evaluation is classification,namely,to identify which one is not overdue or may be overdue.As an excellent classification tool,artificial neural network does not require certain preconditions or assumptions for input and output before modeling.It can be self-learning and self-adaptive,with strong nonlinear mapping ability and fault tolerance mechanism.So,it is a powerful tool to solve the credit problem.The BP neural network is one of the most mature and widely used models,When using neural networks in practical tasks,most of them are trained using BP algorithm.Therefore,based on the theoretical knowledge,this paper uses R language to clean the two data sets and establish a BP neural network model.whit appropriate variables.The established model was evaluated by the criteria of AUC value,etc,and the data set was predicted with better fitting results.