A Big Data-Based Method for Overdue Prediction of Internet Finance User Loans

Abstract

The present invention claims to protect a method for predicting overdue loans of Internet financial users based on big data, including: 101 data preprocessing operations: preprocessing operations based on user behavior, user basic information, and credit score data; Perform division operation; 103 expand the training set according to the confrontation network generation model; 104 perform feature engineering construction operation according to user behavior, user basic information, and credit score data; 105 establish 4 machine learning models, and perform linear regression model fusion operation; 106 pass The model foundation has been established, and artificial threshold rules are set according to the user's basic information to predict whether the user will default. The invention uses big data to realize the transformation of Internet financial risk institutions from the traditional means of "blocking after the event" to "pre-identification" to identify users with high defaults.

Description

technical field [0001] The invention belongs to the technical fields of machine learning, big data processing, deep learning, and artificial intelligence, in particular, a data division method, data expansion, feature construction, model design, artificial rules, etc. of Internet financial user loan overdue prediction data. Background technique [0002] With the rapid development of the Internet, as more and more real economies carry out "Internet +" and the vigorous development of the original Internet industry, a large amount of funds and businesses are realized on the basis of the Internet, and the technology to ensure the flow of these funds and businesses is also As it matures day by day, lending, which is one of the normal financial needs, naturally appears in the network economy. Difficulties of traditional banks: The traditional banking system has been unable to fully solve the needs and difficulties of residents in the field of lending. In 2015, the transaction scal...

AI Technical Summary

Problems solved by technology

[0004] At present, the risk prediction of defaulting users in China adopts the form of manual approval operation, which is inefficient and faces a great risk of default. It is impossible to carry out risk classification management, which affects the ability and flexibility of risk control.

Images