A credit scoring card development method based on machine learning

Abstract

The invention discloses a scoring card development method based on machine learning. The scoring card development method comprises the following steps: (1) defining a label of a target user accordingto vinage analysis; (2) integrating various data sources to obtain final data; (3) carrying out exploratory analysis and data cleaning on the data; (4) using an optimized chi-square binning method tobind the cleaned data; (5) carrying out variable screening on the box-separated variables; (6) constructing a logistic regression model; (7) evaluating the model; And (8) converting the model output target user default probability into a score. According to the method, machine learning, vinage analysis and a logistic regression model are utilized, and the problem is converted from manual solutionto machine solution according to the difficulties of low manual efficiency, difficulty in auditing and the like in the big data era.

Description

technical field [0001] The present invention relates to a kind of Internet finance, machine learning, vintage analysis, logistics regression model, computer application field, what especially relate to is a kind of credit scoring card development method based on machine learning; Background technique [0002] With the rapid development of credit scoring models and the credit industry, there are various methods for building models, from the traditional statistical regression method at the beginning to the emerging deep learning algorithm today, and the application of the model has gradually shifted from predicting the probability of default to various aspects of credit. Life cycle penetration, such as scoring A card, B card after loan, and subsequent C card. However, the scorecards of general financial companies are still traditional expert-type scorecards, which rely on experienced experts to formulate rules to distinguish good and bad users. This method is still effective w...

AI Technical Summary

Problems solved by technology

[0003] In order to overcome the deficiencies of the prior art, the present invention proposes a credit score card development method based on machine learning, using machine learning, vintage analysis, and logistic regression models, aiming at the difficulties of low labor efficiency and difficulty in auditing in the era of big data, and solves the problem Switching from relying on manual solutions to machine solutions

Images