Design And Implementation Of P2P Financial Risk Control System Based On Dig Data

With the transition of consumption structure from subsistence to development and quality consumption,China's consumer finance is bursting.The booming P2P lending industry,which is the representative of Internet consumer finance,has made loans more convenient and faster,and it has also boosted the level of social consumption.However,the current demand for loan shows the trend of miniaturization and rapidness,and the mode of P2P lending is also facing challenges.The credit sources of existing loan products mainly rely on scraping user credit reports,grabbing user credit card bills.There are two major drawbacks of this type of approach.One is that the acquisition is not timely,and the second is the user's credit reports and credit card usage rate.The coverage rate of the information does not provide services for every potential user.At the same time,the popularity of mobile payments in China provides the basis for a new loan model.The P2P financial risk control system is mainly established in response to the demand for online loans under the new situation.It comprehensively analyzes the user's network consumption,mobile payment,and mobile communication information,and provides decisions for the system's upstream and downstream lending and entry processes.Efficient and high-yield online loan fulfillment tasks.The system is essentially a RestFul service provider based on Web.py.At the same time,the training and development of the model and the storage of data are based on the Hadoop ecology,which ensures the storage of massive data and the speed of data operations.The main task of this system is to provide the risk decisions for the entire product line,including whether or not to lend money,lending amounts,and interest rates.This system mainly analyzes the users' Alipay,JD data,and communication records.These original data was derived to features in feature engineering based on the Hadoop platform.After that,the credit model was trained,and the most accurate Xgboost-L credit model was selected through comparison of Logistic Regression and Xgboost models,and then verify the accuracy of the product through A/B test.The author is responsible for the design and implementation of product anti-fraud verification process,as well as the training and development of feature engineering and new models,and participate in the implementation of web services for risk control systems.The system has passed the company's online small flow test,the credit model and the fraudulent process results has met the expected results,and the feature project is successfully reused in the after products.