统计与决策2025，Vol.41Issue(11)：68-73,6.DOI:10.13546/j.cnki.tjyjc.2025.11.011
基于XGBoost算法的返贫预警算法探讨
Discussion on Return-to-poverty Early Warning Algorithm Based on the XGBoost Algorithm
摘要
Abstract
With the complete eradication of rural poverty in China in 2020,effectively connecting the consolidation and ex-pansion of poverty alleviation achievements with rural revitalization has become the new task in the new era.Against this back-drop,the return-to-poverty warning mechanism is crucial for preventing the recurrence of poverty and promoting the revitalization of impoverished areas.Based on the China Family Panel Studies(CFPS)data from 2014 to 2020,this paper employs multi-dimen-sional poverty criteria to comprehensively evaluate the application potential and predictive accuracy of the XGBoost algorithm in poverty recurrence warning.The performance of XGBoost(XGB)is systematically compared with that of Decision Tree(DT),KNN algorithm(KNN),Neural Network(NN),and Logistic Regression(LR)to identify the most suitable algorithm.Additionally,the Shapley decomposition method is applied to identify key influencing factors of poverty recurrence under varying poverty standards,addressing the interpretability challenges inherent in machine learning models.The findings indicate that the XGBoost algorithm exhibits robust predictive capability,achieving high accuracy rates in poverty recurrence warnings across different criteria,there-by demonstrating superior performance as an early warning tool.Meanwhile,the Shapley decomposition method can also effective-ly identify important feature variables,providing a reference basis for predicting poverty recurrence and formulating targeted coun-termeasures.关键词
返贫预警算法/乡村振兴/集成学习/XGBoost算法/Shapley分解Key words
return-to-poverty early warning algorithm/rural revitalization/ensemble learning/XGBoost algorithm/Shapley decomposition分类
管理科学引用本文复制引用
彭乔依,仲黍林,尹娟燕,曾会锋..基于XGBoost算法的返贫预警算法探讨[J].统计与决策,2025,41(11):68-73,6.基金项目
国家社会科学基金重大项目(20&ZD132) （20&ZD132）
