Credit default prediction method and system based on optimized Boruta and XGBoost

Abstract

The invention provides a credit default prediction method and system based on optimized Boruta and XGBoost. The method comprises the steps: obtaining an original data set, carrying out the data preprocessing and feature combination operation of all files in the original data set, and obtaining a credit combination data set; based on a Boruta-XGB feature selection algorithm, performing feature selection on the credit combination data set to obtain a credit data set including an optimal feature subset; based on a credit data set obtained by feature selection, optimizing parameters of an XGBoost algorithm by using a genetic algorithm, and constructing and generating a Boruta-GA-XGB model; and on the basis of the Boruta-GA-XGB model, carrying out default prediction on the to-be-predicted user credit information to obtain a prediction result. All features related to dependent variables are mined by adopting an optimized feature selection algorithm Boruta-XGB, parameters of an XGBoost algorithm are optimized by using a genetic algorithm, and a Boruta-GA-XGB model is constructed to predict user credit default. Therefore, the problem that credit default prediction cannot be optimal due to too many features and too many model parameters in the prior art is solved fundamentally.

Description

technical field [0001] The invention belongs to the field of information technology, and in particular relates to a credit default prediction method and system for optimizing Boruta and XGBoost. Background technique [0002] At present, the domestic consumer finance industry is at a turning point from brutal growth to stable development, and has not yet formed a mature and sound risk control system that adapts to China's national conditions. In order to stabilize the operation of the platform, protect the interests of consumers, and ensure the steady and upward development of my country's consumer finance industry, it is necessary to vigorously strengthen the risk control capabilities of consumer financial institutions, especially the risk control capabilities of the pre-loan review process. If there is a risk control loophole, It will have adverse effects on the economy and society. [0003] Pre-loan review is the most important link in credit risk control. The pre-loan re...

AI Technical Summary

Problems solved by technology

The present invention uses the optimized Boruta feature selection algorithm Boruta-XGB to mine all the features related to the dependent variable, and establishes a good feature system; based on the data set after feature selection, the genetic algorithm is used to optimize the parameters of the XGBoost algorithm, and the Boruta-XGB algorithm is constructed. GA-XGB model, and use the Boruta-GA-XGB model to predict the user's credit default, thus solving the root cause of the problem that the existing excessive number of features and too many parameters of the classification model lead to the failure of optimal credit default prediction

Images