计算机工程与应用2026，Vol.62Issue(17)：35-51,17.DOI:10.3778/j.issn.1002-8331.2507-0279
交互作用特征选择研究综述
Review of Research on Interaction Feature Selection
摘要
Abstract
Feature selection plays a crucial role in enhancing model performance,reducing computational cost,and imp-roving interpretability.Conventional methods mainly assess the relevance between features and target variables,often overlooking their potential interactions,which may cause the omission of key feature combinations.Therefore,researchers have proposed interaction-aware feature selection approaches.This paper systematically summarizes and compares main-stream modeling strategies from five perspectives:interaction order,representation form,structural relation,dependency type,and modeling strategy.It further introduces an"ideal-property"evaluation framework for interaction metrics and provides a comprehensive comparison of representative indicators.Finally,the paper discusses existing challenges and out-line future research directions,offering methodological insights for subsequent studies.关键词
特征选择/特征交互/方法分类/度量标准Key words
feature selection/feature interaction/method classification/measurement standards分类
信息技术与安全科学引用本文复制引用
章祉瑶,聂斌,郑水飞..交互作用特征选择研究综述[J].计算机工程与应用,2026,62(17):35-51,17.基金项目
国家自然科学基金(82260849) （82260849）
江西中医药大学校级科技创新团队发展计划(CXTD22015) （CXTD22015）
江西中医药大学2024年度学科建设经费(2024jzzdxkx026). （2024jzzdxkx026）
