统计与决策2023，Vol.39Issue(23)：28-33,6.DOI:10.13546/j.cnki.tjyjc.2023.23.005
集成学习方法的应用与比较
Application and Comparison of Ensemble Learning Methods
摘要
Abstract
Based on three representative ensemble learning methods such as decision tree and bagging,random forest and gradient boosting tree,this paper establishes prediction models for cumulative claim amount of commercial auto insurance and compulsory insurance,respectively,and introduces the evaluation indexes of composite mean error and stability to compare the prediction effects of different models.The results show that on all data sets,the gradient boosting tree model has strong prediction ability and high stability.The optimal prediction model is further used to prove that the ranking of important risk variables affect-ing different types of insurance is different,and that the effect of the same risk variable on different types of insurance is also dif-ferent.关键词
集成学习/梯度提升树/汽车保险/风险变量重要性Key words
ensemble learning/gradient boosting tree/auto insurance/importance of risk variables分类
管理科学引用本文复制引用
成佩,孟勇..集成学习方法的应用与比较[J].统计与决策,2023,39(23):28-33,6.基金项目
国家社会科学基金青年项目(22CGL071) （22CGL071）
山西省高等学校哲学社会科学基金项目(2022W065) （2022W065）
山西财经大学青年科研基金项目(QN-202018) （QN-202018）
