Research On The Risk Control Model Towards The Debit And Credit Data

With the rapid development of Internet technology,Internet financial industry emerges,such as financial wealth management and online payment.These financial products are rapidly developing with the help of Internet technology and mobile communication technology.Although the reliable Internet technologies provide a strong guarantee for the common operations,the risk control systems of Internet finance still exist in standardization,accuracy,and legitimacy compared with the traditional financial risk control system.These problems have some uncertain financial risks,which not only affect the security of the company's assets,but also influence the prosperity of the country and society.To reduce financial risks,we need a healthy,stable,standardized,and accurate risk control system.As the latest business development model,Internet finance is also a trend of future development.The traditional risk control model mainly analyzes the proportion of various indicators of the risk through people's experience and judges the user's risk.It is feasible under the condition that the users' demands are few and the product is single.But now,more and more products have emerged.At the same time,people's demands for various types of products are constantly rising.The traditional financial risk control models are no longer fit for this situation.This dissertation combines traditional finance with the advanced technologies to help reduce labor costs and time costs,monitor financials in real time,reduce human factors,and increase work efficiency.Using various data mining algorithms,it can quickly and better adapt to the corresponding financial risk control scenarios.The main research work of this dissertation is as follows:(1)Introduce the special feature engineering method for the loan data,introduces the loan data cleaning,feature construction process,and how to select the characteristics of the loan data.After the feature engineering loan data is used to train the corresponding models on the existing five algorithms,the borrowing data is analyzed on the accuracy of these five models.Finally,the five models are weighted and integrated to create a set of ensemble models.This model has the advantages of an integrated wind control model.The evaluation indicators ROC and AUC is used to evaluate the efficiencies of various models.(2)For the traditional rule model,different products correspond to different rules.By studying the architecture of the multiple-integrity weighted model,business process and implementation process are describes.Through the comparison of multivariate integration models with other algorithms,the advantages of multivariate integration weighted models are demonstrated.