Research On The Forecast Of Consumer Loan Capital Flow On Internet Financial Platform

With the rapid development of digital finance and the rapid expansion of online funds,in order to meet the consumer loan needs of users with different qualifications,the threshold of consumer loan products launched by various Internet financial platforms has gradually lowered,resulting in increased risk differences between users,which has brought great challenges to fund management.In this context,this paper conducts research on the prediction of loan capital flow.This paper takes the consumer loan product A launched by an Internet financial platform as an example,takes its platform lending and user repayment capital flow as the research object,and builds a model to predict its loan capital flow.The main work is as follows:（1）Through drawing analysis,it is found that There are differences in the fund flow trends of platform lending and user repayment,which need to be modeled and forecasted separately.（2）Based on the characteristics of capital flow,in addition to building the characteristics of financial and time dimensions,it also crawled related products from Baidu and Weibo.（3）Establish LSTM,GRU,and Wave Net models for prediction research.During the research process,it was found that the Wave Net model has excellent prediction accuracy,model stability and convergence speed.（4）Introduce a two-layer attention mechanism into the simple model,and propose BI-AM-LSTM,BI-AM-GRU,BI-AM-Wave Net models,and the results show that adding a two-layer attention mechanism It can effectively improve the prediction accuracy.（5）In order to further improve the prediction accuracy,the BI-AM-Wave Net model is fused with the LSTM and GRU models,respectively,and the BAWL and BAWG models are proposed.The performance of prediction accuracy has been improved to a certain extent,among which,the prediction effect of the BAWG model is better.（6）Aiming at the problem of insufficient information extraction,the CEEMDAN decomposition method is combined with the BAWG model,and the CBAWG model is proposed.The results show that it has a further improvement in the prediction accuracy,but the improvement is less.In this paper,the prediction of the loan capital flow of product A has a good performance in terms of accuracy and trend grasp,which verifies the feasibility and effectiveness of the proposed model in predicting the capital flow.