Article(id=1327018481364611924, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1327018256403117010, articleNumber=null, orderNo=null, doi=10.7544/issn1000-1239.202550306, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1746547200000, receivedDateStr=2025-05-07, revisedDate=1751990400000, revisedDateStr=2025-07-09, acceptedDate=null, acceptedDateStr=null, onlineDate=1794315460768, onlineDateStr=2026-11-10, pubDate=1756656000000, pubDateStr=2025-09-01, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1794315460768, onlineIssueDateStr=2026-11-10, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1794315460768, creator=13701087609, updateTime=1794315460768, updator=13701087609, issue=Issue{id=1327018256403117010, tenantId=1146029695717560320, journalId=1324759914146529478, year='2025', volume='62', issue='9', pageStart='2125', pageEnd='2395', issueExtLink='null', onlineDate='null', pubDate='1756656000000', pubDateStr='2025-09-01', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1794315407133, creator='13701087609', updateTime=1794376916348, updator='13701087609', preIssue=null, nextIssue=null, articleTotal=null, ext={EN=IssueExt(id=1327276245596876892, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1327018256403117010, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1327276245596876893, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1327018256403117010, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null, downloadFileDto=null}, startPage=2127, endPage=2138, ext={EN=ArticleExt(id=1327018481578521429, articleId=1327018481364611924, tenantId=1146029695717560320, journalId=1324759914146529478, language=EN, title=Reinforcement Learning-Based Feature Generation Algorithm for Scientific Data, columnId=1327018257179063251, journalTitle=Journal of Computer Research and Development, columnName=Special Issue on Advances and Applications of Data-Centric AI, runingTitle=null, highlight=null, articleAbstract=

Feature generation (FG) aims to enhance the prediction potential of original data by constructing high-order feature combinations and removing redundant features. It is a key preprocessing step for tabular scientific data to improve downstream machine-learning model performance. Traditional methods face the following two challenges when dealing with the feature generation of scientific data: First, the effective construction of high-order feature combinations in scientific data necessitates profound and extensive domain-specific expertise. Secondly, as the order of feature combinations increases, the search space expands exponentially, imposing prohibitive human labor consumption. Advancements in the data-centric artificial intelligence (DCAI) paradigm have opened novel avenues for automating feature generation processes. Inspired by that, this paper revisits the conventional feature generation workflow and proposes the multi-agent feature generation (MAFG) framework. Specifically, in the iterative exploration stage, multi-agents will construct mathematical transformation equations collaboratively, synthesize and identify feature combinations exhibiting high information content, and leverage a reinforcement learning mechanism to evolve their strategies. Upon completing the exploration phase, MAFG integrates the large language models (LLMs) to interpretatively evaluate the generated features of each significant model performance breakthrough. Experimental results and case studies consistently demonstrate that MAFG framework effectively automates the feature generation process and significantly enhances various downstream scientific data mining tasks.

, authors=Meng Xiao1, Junfeng Zhou2, Yuanchun Zhou1, 2, *, authorsList=Meng Xiao, Junfeng Zhou, Yuanchun Zhou, authorCompany=null, correspAuthors=Yuanchun Zhou, authorNote=

born in 1995. PhD, assistant professor. Member of CCF. His main research interests include AI4Science and data-centric AI

born in 2002. Undergraduate. His main research interest includes data mining

born in 1975. PhD, professor, PhD supervisor. Senior member of CCF. His main research interests include scientific data and data intelligence

, correspAuthorsNote=null, copyrightStatement=Copyright © 2025 Journal of Computer Research and Development. All rights reserved., copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, fund=null), CN=ArticleExt(id=1327018483382072169, articleId=1327018481364611924, tenantId=1146029695717560320, journalId=1324759914146529478, language=CN, title=基于强化学习的科学数据特征生成算法, columnId=1327018257363612629, journalTitle=计算机研究与发展, columnName=数据为中心的AI前沿进展及应用, runingTitle=null, highlight=null, articleAbstract=

特征生成(feature generation, FG)的目标是通过构建高阶特征组合并去除冗余特征,提升原始数据的预测潜力. 针对表格型科学数据,特征生成是提高下游机器学习模型性能的重要预处理环节. 然而,传统方法在处理科学数据特征生成问题时面临以下两方面挑战:首先,针对科学数据生成有效的高阶特征组合通常需要深入且广泛的领域知识;其次,随着特征组合阶数的增加,组合搜索空间呈指数级扩张,导致人工探索成本过高. 近年来,以数据为核心的人工智能(data-centric artificial intelligence, DCAI)范式的兴起为自动化特征生成过程提供了新的可能性. 受此启发,重新审视了传统特征生成工作流程,并提出了一种多智能体特征生成(multi-agent feature generation, MAFG)框架. 具体而言,在迭代探索阶段,多个智能体协作构建数学变换算式,识别并合成具有高信息含量的特征组合,最后通过强化学习机制实现策略的自适应演化. 探索阶段结束后,MAFG框架引入大语言模型(large language models, LLMs),针对探索过程中的每个关键模型性能突破点,解释性评估该步骤新生成的特征. 实验结果和具体案例研究表明,MAFG框架能够有效地实现特征生成过程的自动化,并能显著提升下游多个科学数据挖掘任务的效果.

, authors=肖濛1, 周骏丰2, 周园春1, 2, *, authorsList=肖濛, 周骏丰, 周园春, authorCompany=null, correspAuthors=周园春, authorNote=

1995年生. 博士,助理研究员. CCF会员. 主要研究方向为科学智能、以数据为核心的人工智能技术

2002年生. 本科生. 主要研究方向为数据挖掘

1975年生. 博士,研究员,博士生导师. CCF高级会员. 主要研究方向为科学数据与数据智能

, correspAuthorsNote=
周园春()
, copyrightStatement=版权所有 © 2025 计算机研究与发展编辑部, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=nb6xdSX5Yb7F3LbGkHMeIQ==, magXml=qOvpb+HpdHjTMyd23NwF6Q==, pdfUrl=null, pdf=uXZQw0d4RTltyQDeJqas9g==, pdfFileSize=1210197, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=z7mKpPIsOks/Sp0zBZFRyQ==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=XZfuHTDHWNIXrJaJn51gRg==, mapNumber=null, fund=null)}, authors=[Author(id=1327018485365977969, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=shaow@cnic.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327018485470835571, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, authorId=1327018485365977969, language=EN, stringName=Meng Xiao, firstName=Meng, middleName=null, lastName=Xiao, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1Computer Network Information Center, Chinese Academy of Sciences, Beijing 100190, bio={"content":"

born in 1995. PhD, assistant professor. Member of CCF. His main research interests include AI4Science and data-centric AI

"}, bioImg=null, bioContent=

born in 1995. PhD, assistant professor. Member of CCF. His main research interests include AI4Science and data-centric AI

, aboutCorrespAuthor=null), CN=AuthorExt(id=1327018485550527348, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, authorId=1327018485365977969, language=CN, stringName=肖濛, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1中国科学院计算机网络信息中心 北京 100190, bio={"img":"JoUPyfGRDgczEV1mKvjZ4Q==","content":"

1995年生. 博士,助理研究员. CCF会员. 主要研究方向为科学智能、以数据为核心的人工智能技术

"}, bioImg=JoUPyfGRDgczEV1mKvjZ4Q==, bioContent=

1995年生. 博士,助理研究员. CCF会员. 主要研究方向为科学智能、以数据为核心的人工智能技术

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327018483600175978, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, xref=1, ext=[AuthorCompanyExt(id=1327018483608564587, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018483600175978, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Computer Network Information Center, Chinese Academy of Sciences, Beijing 100190), AuthorCompanyExt(id=1327018483612758892, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018483600175978, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1中国科学院计算机网络信息中心 北京 100190)])]), Author(id=1327018485638607734, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327018485730882424, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, authorId=1327018485638607734, language=EN, stringName=Junfeng Zhou, firstName=Junfeng, middleName=null, lastName=Zhou, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2University of Chinese Academy of Sciences, Beijing 100049, bio={"content":"

born in 2002. Undergraduate. His main research interest includes data mining

"}, bioImg=null, bioContent=

born in 2002. Undergraduate. His main research interest includes data mining

, aboutCorrespAuthor=null), CN=AuthorExt(id=1327018485806379897, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, authorId=1327018485638607734, language=CN, stringName=周骏丰, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2中国科学院大学 北京 100049, bio={"img":"zQt8Elka1DFf4S5xRqMJ5w==","content":"

2002年生. 本科生. 主要研究方向为数据挖掘

"}, bioImg=zQt8Elka1DFf4S5xRqMJ5w==, bioContent=

2002年生. 本科生. 主要研究方向为数据挖掘

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327018485282091885, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, xref=2, ext=[AuthorCompanyExt(id=1327018485286286190, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018485282091885, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2University of Chinese Academy of Sciences, Beijing 100049), AuthorCompanyExt(id=1327018485303063407, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018485282091885, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2中国科学院大学 北京 100049)])]), Author(id=1327018485886071675, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=zyc@cnic.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1327018485986734974, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, authorId=1327018485886071675, language=EN, stringName=Yuanchun Zhou, firstName=Yuanchun, middleName=null, lastName=Zhou, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, *, address=1Computer Network Information Center, Chinese Academy of Sciences, Beijing 100190
2University of Chinese Academy of Sciences, Beijing 100049, bio={"content":"

born in 1975. PhD, professor, PhD supervisor. Senior member of CCF. His main research interests include scientific data and data intelligence

"}, bioImg=null, bioContent=

born in 1975. PhD, professor, PhD supervisor. Senior member of CCF. His main research interests include scientific data and data intelligence

, aboutCorrespAuthor=null), CN=AuthorExt(id=1327018486058038143, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, authorId=1327018485886071675, language=CN, stringName=周园春, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, *, address=1中国科学院计算机网络信息中心 北京 100190
2中国科学院大学 北京 100049, bio={"img":"bLukmfVLtoW5nY8TNPKQ+A==","content":"

1975年生. 博士,研究员,博士生导师. CCF高级会员. 主要研究方向为科学数据与数据智能

"}, bioImg=bLukmfVLtoW5nY8TNPKQ+A==, bioContent=

1975年生. 博士,研究员,博士生导师. CCF高级会员. 主要研究方向为科学数据与数据智能

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327018483600175978, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, xref=1, ext=[AuthorCompanyExt(id=1327018483608564587, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018483600175978, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Computer Network Information Center, Chinese Academy of Sciences, Beijing 100190), AuthorCompanyExt(id=1327018483612758892, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018483600175978, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1中国科学院计算机网络信息中心 北京 100190)]), AuthorCompany(id=1327018485282091885, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, xref=2, ext=[AuthorCompanyExt(id=1327018485286286190, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018485282091885, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2University of Chinese Academy of Sciences, Beijing 100049), AuthorCompanyExt(id=1327018485303063407, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018485282091885, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2中国科学院大学 北京 100049)])])], keywords=[Keyword(id=1327018486179672960, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, orderNo=1, keyword=deep reinforcement learning), Keyword(id=1327018486255170433, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, orderNo=2, keyword=multi-agent), Keyword(id=1327018486330667906, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, orderNo=3, keyword=data-centric AI), Keyword(id=1327018486397776771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, orderNo=4, keyword=feature generation), Keyword(id=1327018486473274244, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, orderNo=5, keyword=scientific data), Keyword(id=1327018486573937541, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, orderNo=1, keyword=深度强化学习), Keyword(id=1327018486645240710, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, orderNo=2, keyword=多智能体), Keyword(id=1327018486716543879, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, orderNo=3, keyword=以数据为核心的人工智能), Keyword(id=1327018486779458440, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, orderNo=4, keyword=特征生成), Keyword(id=1327018486854955913, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, orderNo=5, keyword=科学数据)], refs=[Reference(id=1327018491539993519, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=1, rfOrder=0, authorNames=null, journalName=null, refType=null, unstructuredReference=Zha Daochen, Bhat Z P, Lai K H, et al. Data-centric artificial intelligence: A survey[J]. ACM Computing Surveys, 2025, 57(5): 1−42, articleTitle=null, refAbstract=null), Reference(id=1327018491602908080, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=1, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Dongjie, Xiao Meng, Wu Min, et al. Reinforcement-enhanced autoregressive feature transformation: Gradient-steered search in continuous space for postfix expressions[C]// Proc of Conf on Neural Information Processing Systems. New Orleans, LA: NeurIPS, 2023: 43563−43578, articleTitle=null, refAbstract=null), Reference(id=1327018491665822641, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=2, authorNames=null, journalName=null, refType=null, unstructuredReference=李明阳, 许可儿, 宋志强, 等. 多智能体强化学习算法研究综述[J]. 计算机科学与探索, 2024, 18(8): 1979−1997, articleTitle=null, refAbstract=null), Reference(id=1327018491720348594, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=3, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Mingyang, Xu Keer, Song Zhiqiang, et al. Review of research on multi-agent reinforcement learning algorithms[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(8): 1979−1997 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1327018491808428979, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=4, authorNames=null, journalName=null, refType=null, unstructuredReference=陈人龙, 陈嘉礼, 李善琦, 等. 多智能体强化学习方法综述[J]. 信息对抗技术, 2024, 3(1): 18−32, articleTitle=null, refAbstract=null), Reference(id=1327018491871343540, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=5, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Renlong, Chen Jiali, Li Shanqi, et al. A survey of multi-agent reinforcement learning methods[J]. Information Countermeasure Technology, 2024, 3(1): 18−32 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1327018491959423925, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=6, authorNames=null, journalName=null, refType=null, unstructuredReference=He Tianqi, Huang Xiaohan, Du Yi, et al. FastFT: Accelerating reinforced feature transformation via advanced exploration strategies[C]//Proc of 2025 IEEE 41st Int Conf on Data Engineering (ICDE). Los Alamitos, CA: IEEE Computer Society, 2025: 4120−4133, articleTitle=null, refAbstract=null), Reference(id=1327018492022338486, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=6, rfOrder=7, authorNames=null, journalName=null, refType=null, unstructuredReference=Katz G, Shin E C R, Song D. ExploreKit: Automatic feature generation and selection[C]//Proc of 2016 IEEE 16th Int Conf on Data Mining. Piscataway, NJ: IEEE, 2016: 979−984, articleTitle=null, refAbstract=null), Reference(id=1327018492089447351, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=7, rfOrder=8, authorNames=null, journalName=null, refType=null, unstructuredReference=Dor O, Reich Y. Strengthening learning algorithms by feature discovery[J]. Information Sciences, 2012, 189: 176−190, articleTitle=null, refAbstract=null), Reference(id=1327018492148167608, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=8, rfOrder=9, authorNames=null, journalName=null, refType=null, unstructuredReference=Xiao Meng, Wang Dongjie, Wu Min, et al. Traceable group-wise self-optimizing feature transformation learning: A dual optimization perspective[J]. ACM Transactions on Knowledge Discovery from Data, 2024, 18(4): 1−22, articleTitle=null, refAbstract=null), Reference(id=1327018492232053690, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=9, rfOrder=10, authorNames=null, journalName=null, refType=null, unstructuredReference=Khurana U, Samulowitz H, Turaga D. Feature engineering for predictive modeling using reinforcement learning[C/OL]//Proc of the AAAI Conf on Artificial Intelligence. Menlo Park, CA: AAAI, 2018[2025-07-09]. https://ojs.aaai.org/index.php/AAAI/article/view/11678/11537., articleTitle=null, refAbstract=null), Reference(id=1327018492311745467, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=11, authorNames=null, journalName=null, refType=null, unstructuredReference=Xiao Meng, Wang Dongjie, Wu Min, et al. Traceable automatic feature transformation via cascading actor-critic agents[C]//Proc of the 2023 SIAM Int Conf on Data Mining (SDM). Philadelphia, PA: Society for Industrial and Applied Mathematics, 2023: 775−783, articleTitle=null, refAbstract=null), Reference(id=1327018492391437244, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=11, rfOrder=12, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhu Guanghui, Xu Zhuoer, Yuan Chunfeng, et al. DIFER: Differentiable automated feature engineering[C]// Proc of the 1st Int Conf on Automated Machine Learning. New York: PMLR, 2022, 188: 17/1−17, articleTitle=null, refAbstract=null), Reference(id=1327018492483711933, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=12, rfOrder=13, authorNames=null, journalName=null, refType=null, unstructuredReference=Bellman R. Dynamic programming[J]. Science, 1966, 153(3731): 34−37, articleTitle=null, refAbstract=null), Reference(id=1327018492550820798, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=13, rfOrder=14, authorNames=null, journalName=null, refType=null, unstructuredReference=马琦, 刘杨, 吴贤生, 等. 基于价值迭代算法的最优渗透路径发现. 计算机系统应用, 2023, 32(12): 197−204, articleTitle=null, refAbstract=null), Reference(id=1327018494211765183, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=13, rfOrder=15, authorNames=null, journalName=null, refType=null, unstructuredReference=Ma Qi, Liu Yang, Wu Xiansheng, et al. Optimal penetration path discovery based on value iterative algorithm[J]. Computer Systems Applications, 2023, 32(12): 197−204 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1327018494354371520, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=14, rfOrder=16, authorNames=null, journalName=null, refType=null, unstructuredReference=Ding Zihan, Huang Yanhua, Yuan Hang, et al. Introduction to reinforcement learning[J]. Deep Reinforcement Learning: Fundamentals, Research and Applications, 2020: 47−123, articleTitle=null, refAbstract=null), Reference(id=1327018494496977857, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=17, authorNames=null, journalName=null, refType=null, unstructuredReference=Watkins C J C H, Dayan P. Q-learning[J]. Machine Learning, 1992, 8: 279−292, articleTitle=null, refAbstract=null), Reference(id=1327018494576669634, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=16, rfOrder=18, authorNames=null, journalName=null, refType=null, unstructuredReference=Mnih V, Kavukcuoglu K, Silver D, et al. Playing Atari with deep reinforcement learning[J]. arXiv preprint, arXiv: 1312.5602, 2013, articleTitle=null, refAbstract=null), Reference(id=1327018494664750019, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=19, authorNames=null, journalName=null, refType=null, unstructuredReference=Sutton R S, McAllester D, Singh S, et al. Policy gradient methods for reinforcement learning with function approximation[C/OL]// Proc of Conf on Neural Information Processing Systems. Denver, CO, USA: NeurIPS, 1999[2025-07-09]. https://proceedings.neurips.cc/paper/1713-policy-gradient-methods-for-reinforcement-learning-with-function-approximat., articleTitle=null, refAbstract=null), Reference(id=1327018494736053188, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=18, rfOrder=20, authorNames=null, journalName=null, refType=null, unstructuredReference=Mnih V, Badia A P, Mirza M, et al. Asynchronous methods for deep reinforcement learning[C]//Proc of the 33rd Int Conf on Machine Learning. New York: PMLR, 2016: 1928−1937, articleTitle=null, refAbstract=null), Reference(id=1327018494794773445, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=19, rfOrder=21, authorNames=null, journalName=null, refType=null, unstructuredReference=况立群, 冯利, 韩燮, 等. 基于双深度网络的智能决策系统研究[J]. 计算机技术与发展, 2022, 32(2): 137−142, articleTitle=null, refAbstract=null), Reference(id=1327018494861882310, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=19, rfOrder=22, authorNames=null, journalName=null, refType=null, unstructuredReference=Kuang Liqun, Feng Li, Han Xie, et al. Research on intelligent decision-making system based on double deep Q-Network[J]. Computer Technology and Development, 2022, 32(2): 137−142 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1327018494928991175, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=23, authorNames=null, journalName=null, refType=null, unstructuredReference=张俊杰, 张聪, 赵涵捷. 重复利用状态值的竞争深度网络算法[J]. 计算机工程与应用, 2021, 57(4): 134−140, articleTitle=null, refAbstract=null), Reference(id=1327018495008682952, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=24, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Junjie, Zhang Cong, Zhao Hanjie. Dueling deep Q network algorithm with state value reuse[J]. Computer Engineering and Applications, 2021, 57(4): 134−140 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1327018495075791817, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=21, rfOrder=25, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Yunxia. Data set of hand, foot and mouth disease and main meteorological factors at county level in Shandong Province, China, 2015−2019[DS/OL]. V1. Science Data Bank, 2023[2025-07-09]. https://cstr.cn/31253.11.sciencedb.07251, articleTitle=null, refAbstract=null), Reference(id=1327018495138706378, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=22, rfOrder=26, authorNames=null, journalName=null, refType=null, unstructuredReference=Yu Xiao. Data set of learning engagement and environmental variables of Chinese Primary School children[DS/OL]. V2. Science Data Bank, 2023[2025-07-09]. https://cstr.cn/31253.11.sciencedb.06343, articleTitle=null, refAbstract=null), Reference(id=1327018495226786763, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=23, rfOrder=27, authorNames=null, journalName=null, refType=null, unstructuredReference=Xing Yuping, Xing Hui, Li Miao, et al. Relationship between cystatin C levels and risk of frailty in two Chinese populations aged 45~65 years and 65 years or older[DS/OL]. V1. Science Data Bank, 2023[2025-07-09]. https://cstr.cn/31253.11.sciencedb.o00130.00340, articleTitle=null, refAbstract=null), Reference(id=1327018495298089932, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=24, rfOrder=28, authorNames=null, journalName=null, refType=null, unstructuredReference=Tsou Y L, Lin Y W, Shao H Y, et al. Recombinant adeno-vaccine expressing enterovirus 71-like particles against hand, foot, and mouth disease[J]. PLoS Neglected Tropical Diseases, 2015, 9(4): e0003692, articleTitle=null, refAbstract=null), Reference(id=1327018495361004493, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=25, rfOrder=29, authorNames=null, journalName=null, refType=null, unstructuredReference=Nguyen A T, Grembi J A, Riviere M, et al. Influence of temperature and precipitation on the effectiveness of water, sanitation, and handwashing interventions against childhood diarrheal disease in rural Bangladesh: A reanalysis of the WASH Benefits Bangladesh trial[J]. Environmental Health Perspectives, 2024, 132(4): 047006, articleTitle=null, refAbstract=null), Reference(id=1327018495411336142, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=26, rfOrder=30, authorNames=null, journalName=null, refType=null, unstructuredReference=Andhikaputra G, Sapkota A, Lin Y K, et al. The impact of temperature and precipitation on all-infectious-, bacterial-, and viral-diarrheal disease in Taiwan[J]. Science of The Total Environment, 2023, 862: 160850, articleTitle=null, refAbstract=null), Reference(id=1327018495465862095, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=27, rfOrder=31, authorNames=null, journalName=null, refType=null, unstructuredReference=Wesenberg L, Jansen S, Krieglstein F, et al. The influence of seductive details in learning environments with low and high extrinsic motivation[J]. Learning and Instruction, 2025, 96: 102054, articleTitle=null, refAbstract=null), Reference(id=1327018495524582352, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=28, rfOrder=32, authorNames=null, journalName=null, refType=null, unstructuredReference=Garrote A, Diener M, Hepberger B, et al. Social behavior, academic achievement, language skills, and peer rejection in elementary school classes: The moderating role of teacher feedback[J]. Educational Psychology, 2024, 44(5): 613−631, articleTitle=null, refAbstract=null), Reference(id=1327018495595885521, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=29, rfOrder=33, authorNames=null, journalName=null, refType=null, unstructuredReference=Kong Yangtao. The role of experiential learning on students’ motivation and classroom engagement[J]. Frontiers in Psychology, 2021, 12: 771272, articleTitle=null, refAbstract=null), Reference(id=1327018495675577298, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=30, rfOrder=34, authorNames=null, journalName=null, refType=null, unstructuredReference=Bukabau J B, Yayo E, Gnionsahé A, et al. Performance of creatinine-or cystatin C–based equations to estimate glomerular filtration rate in sub-Saharan African populations[J]. Kidney International, 2019, 95(5): 1181−1189, articleTitle=null, refAbstract=null), Reference(id=1327018495759463379, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=31, rfOrder=35, authorNames=null, journalName=null, refType=null, unstructuredReference=Rudberg S, Rasmussen L M, Bangstad H J, et al. Influence of insertiondeletion polymorphism in the ACE-I gene on the progression of diabetic glomerulopathy in type 1 diabetic patients with microalbuminuria[J]. Diabetes Care, 2000, 23(4): 544−548, articleTitle=null, refAbstract=null), Reference(id=1327018495834960852, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=32, rfOrder=36, authorNames=null, journalName=null, refType=null, unstructuredReference=Lim S K, Kim S M I, Lee E K, et al. Usefulness of cystatin C in measuring renal function in patients with liver cirrhosis[J]. Nephrology Dialysis Transplantation, 2024, 39(Supplement_1): g−f, articleTitle=null, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1327018483600175978, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, xref=1, ext=[AuthorCompanyExt(id=1327018483608564587, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018483600175978, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Computer Network Information Center, Chinese Academy of Sciences, Beijing 100190), AuthorCompanyExt(id=1327018483612758892, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018483600175978, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1中国科学院计算机网络信息中心 北京 100190)]), AuthorCompany(id=1327018485282091885, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, xref=2, ext=[AuthorCompanyExt(id=1327018485286286190, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018485282091885, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2University of Chinese Academy of Sciences, Beijing 100049), AuthorCompanyExt(id=1327018485303063407, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018485282091885, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2中国科学院大学 北京 100049)])], figs=[ArticleFig(id=1327018486993367946, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Figure 1, caption=Framework of MAFG, figureFileSmall=2AJudMBjl0kdn5vEzxWmHQ==, figureFileBig=z7mKpPIsOks/Sp0zBZFRyQ==, tableContent=null), ArticleFig(id=1327018487094031243, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=图1, caption=MAFG架构, figureFileSmall=2AJudMBjl0kdn5vEzxWmHQ==, figureFileBig=z7mKpPIsOks/Sp0zBZFRyQ==, tableContent=null), ArticleFig(id=1327018487295357836, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Figure 2, caption=Comparison of performance changes under different DQN algorithms, figureFileSmall=xFzAMNNqWM3aiE8EwHhtyg==, figureFileBig=ystUdcqi/KHhSoT94gslwg==, tableContent=null), ArticleFig(id=1327018487362466701, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=图2, caption=不同DQN算法数性能变化对比, figureFileSmall=xFzAMNNqWM3aiE8EwHhtyg==, figureFileBig=ystUdcqi/KHhSoT94gslwg==, tableContent=null), ArticleFig(id=1327018487467324302, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Figure 3, caption=Comparison of different feature selection components, figureFileSmall=v7rMMgffXJKzq9SarG9AjQ==, figureFileBig=UvsqQjXV7Z5mCPHFVTWcpA==, tableContent=null), ArticleFig(id=1327018487555404687, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=图3, caption=不同特征选择组件的对比, figureFileSmall=v7rMMgffXJKzq9SarG9AjQ==, figureFileBig=UvsqQjXV7Z5mCPHFVTWcpA==, tableContent=null), ArticleFig(id=1327018487647679376, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Figure 4, caption=Performance variation of the hand-foot-mouth disease meteorological dataset, figureFileSmall=BchJUTufiqZUgBr0VTmcZQ==, figureFileBig=atLMMmt2tuYoZlM3Z0nIbw==, tableContent=null), ArticleFig(id=1327018487714788241, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=图4, caption=手足口病气象数据集性能变化, figureFileSmall=BchJUTufiqZUgBr0VTmcZQ==, figureFileBig=atLMMmt2tuYoZlM3Z0nIbw==, tableContent=null), ArticleFig(id=1327018487802868626, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Figure 5, caption=Performance variation of primary school students’ learning engagement dataset, figureFileSmall=W0tdfTmNtk985nqtE2hktA==, figureFileBig=LaZFAlpYcS63Wg3Vvfps7A==, tableContent=null), ArticleFig(id=1327018487861588883, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=图5, caption=小学生学习投入数据集性能变化, figureFileSmall=W0tdfTmNtk985nqtE2hktA==, figureFileBig=LaZFAlpYcS63Wg3Vvfps7A==, tableContent=null), ArticleFig(id=1327018487941280660, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Figure 6, caption=Performance change of Cystatin C and frailty risk dataset, figureFileSmall=Pxa1s8R3k2Ya7EE3ltScUg==, figureFileBig=CWPEa16ZMmQNxMzSkZez7Q==, tableContent=null), ArticleFig(id=1327018488016778133, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=图6, caption=胱抑素C与衰弱风险数据集性能变化, figureFileSmall=Pxa1s8R3k2Ya7EE3ltScUg==, figureFileBig=CWPEa16ZMmQNxMzSkZez7Q==, tableContent=null), ArticleFig(id=1327018488088081302, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Table 1, caption=Performance Analysis of Hand-foot-mouth Disease Meteorological Dataset, figureFileSmall=null, figureFileBig=null, tableContent=
性能指标原始数据MAFG (本文)变化幅度
1-RAE0.3020.358+18.5%
MAE37.38537.366−0.05%
RMSE5983.2575713.384−4.51%
), ArticleFig(id=1327018489778385815, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=表1, caption=

手足口病气象数据集性能分析

, figureFileSmall=null, figureFileBig=null, tableContent=
性能指标原始数据MAFG (本文)变化幅度
1-RAE0.3020.358+18.5%
MAE37.38537.366−0.05%
RMSE5983.2575713.384−4.51%
), ArticleFig(id=1327018489862271896, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Table 2, caption=Performance Analysis of Primary School Learning Input Dataset, figureFileSmall=null, figureFileBig=null, tableContent=
性能指标原始数据MAFG (本文)变化幅度
1-RAE0.8960.953+6.3%
MAE0.0790.038−51.8%
RMSE0.0130.004−69.2%
), ArticleFig(id=1327018489933575066, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=表2, caption=

小学生学习投入数据集性能分析

, figureFileSmall=null, figureFileBig=null, tableContent=
性能指标原始数据MAFG (本文)变化幅度
1-RAE0.8960.953+6.3%
MAE0.0790.038−51.8%
RMSE0.0130.004−69.2%
), ArticleFig(id=1327018490009072539, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Table 3, caption=Performance Analysis of Cystatin C and Frailty Risk Dataset, figureFileSmall=null, figureFileBig=null, tableContent=
性能指标原始数据MAFG变化幅度
1-RAE0.0020.035+1650%
MAE8.5008.188−3.6%
RMSE122.857119.254−2.9%
), ArticleFig(id=1327018490080375708, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=表3, caption=

胱抑素C与衰弱风险数据集性能分析

, figureFileSmall=null, figureFileBig=null, tableContent=
性能指标原始数据MAFG变化幅度
1-RAE0.0020.035+1650%
MAE8.5008.188−3.6%
RMSE122.857119.254−2.9%
), ArticleFig(id=1327018490151678877, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Table 4, caption=Importance of Original Features of Hand-foot-mouth Disease Meteorological Dataset, figureFileSmall=null, figureFileBig=null, tableContent=
特征重要性
注:“population”表示报告病例所在地区的人口总数;“temperature”表示病例所在地区在报告病例当天的气温;“humidity”表示病例所在地区在报告病例当天的相对湿度;“air pressure”表示病例所在地区在报告病例当天的气压;“precipitation”表示病例所在地区在报告病例当天的降水量;“wind speed”表示病例所在地区在报告病例当天的风速;“sunshine hour”表示该地区当日日照时间.
population0.180
temperature0.286
humidity0.105
air pressure0.069
precipitation0.219
wind speed0.054
sunshine hour0.202
), ArticleFig(id=1327018490218787742, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=表4, caption=

手足口病气象数据集原特征重要性

, figureFileSmall=null, figureFileBig=null, tableContent=
特征重要性
注:“population”表示报告病例所在地区的人口总数;“temperature”表示病例所在地区在报告病例当天的气温;“humidity”表示病例所在地区在报告病例当天的相对湿度;“air pressure”表示病例所在地区在报告病例当天的气压;“precipitation”表示病例所在地区在报告病例当天的降水量;“wind speed”表示病例所在地区在报告病例当天的风速;“sunshine hour”表示该地区当日日照时间.
population0.180
temperature0.286
humidity0.105
air pressure0.069
precipitation0.219
wind speed0.054
sunshine hour0.202
), ArticleFig(id=1327018490285896607, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Table 5, caption=Generated Feature Combinations for Hand-foot-mouth Disease Meteorological Dataset, figureFileSmall=null, figureFileBig=null, tableContent=
特征组合步骤重要性
cube(population)E1-S10.199
stand_scaler(air_pressure)E1-S30.101
temperature × precipitationE2-S40.281
reciprocal(sunshine_hour)E3-S10.067
sigmoid(air_pressure+population)E3-S30.207
precipitation×air_pressureE4-S40.218
tanh(temperature)E5-S20.017
), ArticleFig(id=1327018490365588384, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=表5, caption=

手足口病气象数据集生成特征组合

, figureFileSmall=null, figureFileBig=null, tableContent=
特征组合步骤重要性
cube(population)E1-S10.199
stand_scaler(air_pressure)E1-S30.101
temperature × precipitationE2-S40.281
reciprocal(sunshine_hour)E3-S10.067
sigmoid(air_pressure+population)E3-S30.207
precipitation×air_pressureE4-S40.218
tanh(temperature)E5-S20.017
), ArticleFig(id=1327018490436891553, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Table 6, caption=Discussion Environmental Significance Results of Generated Feature Combination, figureFileSmall=null, figureFileBig=null, tableContent=
特征环境意义文献支持
temperature温度升高加速病毒
复制与媒介生物活动
PLoS NTD指出温度大于25°C时
EV71病毒活性提升3倍①
precipitation降水增加形成积水环
境,促进蚊媒滋生
Environ Health Perspect证实降水
与手足口病传播呈J型曲线关系②
temperature×
precipitation
高温高湿环境产
生协同放大效应
Sci Total Environ发现温雨乘积
项比单因素预测效能高41%③
), ArticleFig(id=1327018490504000418, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=表6, caption=

生成的特征组合环境意义讨论结果

, figureFileSmall=null, figureFileBig=null, tableContent=
特征环境意义文献支持
temperature温度升高加速病毒
复制与媒介生物活动
PLoS NTD指出温度大于25°C时
EV71病毒活性提升3倍①
precipitation降水增加形成积水环
境,促进蚊媒滋生
Environ Health Perspect证实降水
与手足口病传播呈J型曲线关系②
temperature×
precipitation
高温高湿环境产
生协同放大效应
Sci Total Environ发现温雨乘积
项比单因素预测效能高41%③
), ArticleFig(id=1327018490575303587, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Table 7, caption=Original Feature Importance of the Primary School Students’ Learning Engagement Dataset, figureFileSmall=null, figureFileBig=null, tableContent=
特征重要性
注:“gender”表示性别;“T3age”指第3次数据收集时的年龄;“T3grade”指第3次数据收集时的年级;“ZT2PS”指标准化后的积极亲子风格得分;“ZT2TS”指标准化后的积极师生关系得分;“ZT2PR”指标准化后的同伴关系得分;“ZT3vigor”指“活力”维度得分;“ZT3dedication”指“奉献”维度得分;“ZT3absorption”指“专注”维度得分.
gender0.0086
T3age0.0075
T3grade0.0094
ZT2PS0.1784
ZT2TS0.1663
ZT2PR0.1051
ZT3vigor0.1966
ZT3dedication0.1853
ZT3absorption0.1830
), ArticleFig(id=1327018490646606756, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=表7, caption=

小学生学习投入数据集原特征重要性

, figureFileSmall=null, figureFileBig=null, tableContent=
特征重要性
注:“gender”表示性别;“T3age”指第3次数据收集时的年龄;“T3grade”指第3次数据收集时的年级;“ZT2PS”指标准化后的积极亲子风格得分;“ZT2TS”指标准化后的积极师生关系得分;“ZT2PR”指标准化后的同伴关系得分;“ZT3vigor”指“活力”维度得分;“ZT3dedication”指“奉献”维度得分;“ZT3absorption”指“专注”维度得分.
gender0.0086
T3age0.0075
T3grade0.0094
ZT2PS0.1784
ZT2TS0.1663
ZT2PR0.1051
ZT3vigor0.1966
ZT3dedication0.1853
ZT3absorption0.1830
), ArticleFig(id=1327018490734687141, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Table 8, caption=Generated Feature Combinations for the Primary School Students’ Learning Engagement Dataset, figureFileSmall=null, figureFileBig=null, tableContent=
特征组合步骤重要性
ZT2PS ÷ genderE1-S10.0686
ZT2TS × ZT2PRE2-S20.1346
sqrt(ZT3absorption × ZT3dedication)E3-S40.1929
(ZT3dedication3 × ZT2TS2) / ZT3absorptionE3-S50.1742
sqrt(ZT3vigor − ZT3absorption)E4-S40.0932
Sigmoid (ZT2PS + T3age)E5-S10.1270
), ArticleFig(id=1327018490831156134, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=表8, caption=

小学生学习投入数据集生成特征组合

, figureFileSmall=null, figureFileBig=null, tableContent=
特征组合步骤重要性
ZT2PS ÷ genderE1-S10.0686
ZT2TS × ZT2PRE2-S20.1346
sqrt(ZT3absorption × ZT3dedication)E3-S40.1929
(ZT3dedication3 × ZT2TS2) / ZT3absorptionE3-S50.1742
sqrt(ZT3vigor − ZT3absorption)E4-S40.0932
Sigmoid (ZT2PS + T3age)E5-S10.1270
), ArticleFig(id=1327018490915042215, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Table 9, caption=Discussion Results of Generated Features Significance from the Primary School Students’ Learning Engagement Dataset, figureFileSmall=null, figureFileBig=null, tableContent=
特征教育学意义文献支持
ZT3dedication3立方放大“奉献精神”的阈值效应,大于4.2分时产生“心流体验”持续效应Learning and Instruction
证实奉献度3比线性项
更能预测学业成就①
ZT2TS2二次项捕捉师生关系的边际递增效应,优质师生互动对学业的促进作用呈加速趋势Educational Psychology
发现师生关系2的预测效能比线性高29% ②
ZT3absorption分母设计体现“过度专注可能抑制师生互动获益”的调节效应Frontiers in Psychology
报道专注度大于7分时
需平衡师生互动③
), ArticleFig(id=1327018491007316904, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=表9, caption=

小学生学习投入数据集生成特征意义讨论结果

, figureFileSmall=null, figureFileBig=null, tableContent=
特征教育学意义文献支持
ZT3dedication3立方放大“奉献精神”的阈值效应,大于4.2分时产生“心流体验”持续效应Learning and Instruction
证实奉献度3比线性项
更能预测学业成就①
ZT2TS2二次项捕捉师生关系的边际递增效应,优质师生互动对学业的促进作用呈加速趋势Educational Psychology
发现师生关系2的预测效能比线性高29% ②
ZT3absorption分母设计体现“过度专注可能抑制师生互动获益”的调节效应Frontiers in Psychology
报道专注度大于7分时
需平衡师生互动③
), ArticleFig(id=1327018491087008681, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Table 10, caption=Original Feature Importance of Cystatin C and Fateful Risk Dataset, figureFileSmall=null, figureFileBig=null, tableContent=
特征重要性
注:“cystatinc”表示胱抑素C,一种肾功能指标;“newcrp”表示C反应蛋白,一种炎症标志物;“newhbalc”表示糖化血红蛋白,用于评估糖尿病控制情况;“newglu”表示血液中的葡萄糖;“newcho”表示血液中的胆固醇等指标;“newbun”表示血液中的尿素氮;“newcrea”表示血液中的肌酐.
cystatinc0.0299
newcrp0.0038
newhbalc0.0010
newglu0.0094
newcho0.0049
newbun0.0075
newcrea0.0058
), ArticleFig(id=1327018491175089066, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=表10, caption=

胱抑素C与衰弱风险数据集原特征重要性

, figureFileSmall=null, figureFileBig=null, tableContent=
特征重要性
注:“cystatinc”表示胱抑素C,一种肾功能指标;“newcrp”表示C反应蛋白,一种炎症标志物;“newhbalc”表示糖化血红蛋白,用于评估糖尿病控制情况;“newglu”表示血液中的葡萄糖;“newcho”表示血液中的胆固醇等指标;“newbun”表示血液中的尿素氮;“newcrea”表示血液中的肌酐.
cystatinc0.0299
newcrp0.0038
newhbalc0.0010
newglu0.0094
newcho0.0049
newbun0.0075
newcrea0.0058
), ArticleFig(id=1327018491238003627, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Table 11, caption=Generated Feature Combinations for Cystatin C and Frailty Risk Datasets, figureFileSmall=null, figureFileBig=null, tableContent=
特征组合步骤重要性
sqrt(cystatinc)E1-S10.0078
cystatinc3E1-S50.0056
cystatinc × newcrpE2-S30.0132
minmax_scaler(cystatinc)E3-S40.0025
sigmoid (cystatinc)E4-S40.0034
(newcrp2 + newhbalc) × cystatincE5-S30.0093
), ArticleFig(id=1327018491300918188, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=表11, caption=

胱抑素C与衰弱风险数据集生成特征组合

, figureFileSmall=null, figureFileBig=null, tableContent=
特征组合步骤重要性
sqrt(cystatinc)E1-S10.0078
cystatinc3E1-S50.0056
cystatinc × newcrpE2-S30.0132
minmax_scaler(cystatinc)E3-S40.0025
sigmoid (cystatinc)E4-S40.0034
(newcrp2 + newhbalc) × cystatincE5-S30.0093
), ArticleFig(id=1327018491363832749, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=EN, label=Table 12, caption=Discussion Results of Biological Significance of Generated Feature Combinations, figureFileSmall=null, figureFileBig=null, tableContent=
特征教育学意义文献支持
newcrp2CRP的平方项反映炎症的指数级放大效应Kidney International指出CRP2可预测CKD患者心血管事件风险①
newhbalc长期血糖控制指标,高值导致晚期糖基化终产物(AGEs)堆积,加速肾间质纤维化Diabetes Care证实HbA1c与
肾小球硬化面积正相关②
cystatinc肾小球滤过率敏感指标,立方操作可能对应肾单位损失的临界点效应Nephrology Dialysis Transplantation提出cystatinc3可识别早期肾功能非线性衰减③
), ArticleFig(id=1327018491435135918, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, language=CN, label=表12, caption=

生成特征组合生物学意义讨论结果

, figureFileSmall=null, figureFileBig=null, tableContent=
特征教育学意义文献支持
newcrp2CRP的平方项反映炎症的指数级放大效应Kidney International指出CRP2可预测CKD患者心血管事件风险①
newhbalc长期血糖控制指标,高值导致晚期糖基化终产物(AGEs)堆积,加速肾间质纤维化Diabetes Care证实HbA1c与
肾小球硬化面积正相关②
cystatinc肾小球滤过率敏感指标,立方操作可能对应肾单位损失的临界点效应Nephrology Dialysis Transplantation提出cystatinc3可识别早期肾功能非线性衰减③
)], attaches=null, journal=Journal(id=1324759606947315908, delFlag=0, nameCn=计算机研究与发展, nameEn=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, issn=1000-1239, eissn=null, cn=11-1777/TP, coden=null, periodic=0, language=CN, oaType=null, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=k2l8zgqwstuxxevFX2qxVA==, journalPrice=null, startedYear=null, abbrevIsoEn=Journal of Computer Research and Development, journalRemark=null, publicationField=null, createdTime=1793776903141, updatedTime=1793850698612, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=J, firstLetterEn=J, subjectCode=Electronic Information, subjectName=null, subjectCodeEn=Electronic Information, subjectNameEn=null, picCn=k2l8zgqwstuxxevFX2qxVA==, picEn=r6k0Ry+RD+c/VXj5QarmDw==, jcr=null, cjcr=null, exts=[JournalExt(id=1325069127710286688, language=CN, name=计算机研究与发展, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698640, updatedTime=1793850698640, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1325069127777395553, language=EN, name=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698656, updatedTime=1793850698656, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1324759914146529478, websiteList=[Website(id=1325070300139892771, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN, language=CN, createTime=1793850978169, createBy=18614031015, updateTime=1793851040259, updateBy=18614031015, name=计算机研究与发展-中文, tplId=1146099689490845704, title=计算机研究与发展, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070893751349360, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=articleTextType, value=kx, createTime=1793851119697, updateTime=1793851119697, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893730377837, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=banner, value=null, createTime=1793851119692, updateTime=1793851119692, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893772320883, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=grayFlag, value=0, createTime=1793851119702, updateTime=1793851119702, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893721989228, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851119690, updateTime=1793851119690, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893784903797, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=minRunFlag, value=0, createTime=1793851119705, updateTime=1793851119705, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893742960751, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic, createTime=1793851119695, updateTime=1793851119695, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893780709492, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=silenceFlag, value=0, createTime=1793851119704, updateTime=1793851119704, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893738766446, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1793851119694, updateTime=1793851119694, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893755543665, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeColor, value=null, createTime=1793851119698, updateTime=1793851119698, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893763932274, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeStyle, value=null, createTime=1793851119700, updateTime=1793851119700, creator=18614031015, updator=18614031015)]), Website(id=1325070300580294698, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN, language=EN, createTime=1793850978274, createBy=18614031015, updateTime=1793851036638, updateBy=18614031015, name=计算机研究与发展-英文, tplId=1146101810881728533, title=Journal of Computer Research and Development, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070925288321148, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=articleTextType, value=kx, createTime=1793851127216, updateTime=1793851127216, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925263155321, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=banner, value=null, createTime=1793851127210, updateTime=1793851127210, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925309292671, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=grayFlag, value=0, createTime=1793851127221, updateTime=1793851127221, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925254766712, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851127208, updateTime=1793851127208, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925326069889, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=minRunFlag, value=0, createTime=1793851127225, updateTime=1793851127225, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925279932539, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic, createTime=1793851127214, updateTime=1793851127214, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925317681280, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=silenceFlag, value=0, createTime=1793851127223, updateTime=1793851127223, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925271543930, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1793851127212, updateTime=1793851127212, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925292515453, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeColor, value=null, createTime=1793851127217, updateTime=1793851127217, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925300904062, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeStyle, value=null, createTime=1793851127219, updateTime=1793851127219, creator=18614031015, updator=18614031015)])], journalTitle=计算机研究与发展, weixinUrl=null, journalUrl=https://crad.ict.ac.cn/, iacademicId=null, status=1, seqNo=null, journalTitleEn=Journal of Computer Research and Development, journalPhotoCn=k2l8zgqwstuxxevFX2qxVA==, journalPhotoEn=r6k0Ry+RD+c/VXj5QarmDw==, journalFirstLetter=J, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=, provinceCode=null, provinceName=null, collectFlag=false, interPubPlatform=, interPubPlatformUrl=null), detailUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202550306, detailUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/10.7544/issn1000-1239.202550306, pdfUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/PDF/10.7544/issn1000-1239.202550306, pdfUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/PDF/10.7544/issn1000-1239.202550306, aliStartDate=0, aliEndDate=0, collectionFlag=false, citedCount=null, citedUrl=null, previewStatus=0, delFlag=0, hasFullText=1, orderTime=1756656000000, fullTextJson=null, articleText=null, reference=null)
收藏切换
基于强化学习的科学数据特征生成算法
肖濛 1 , 周骏丰 2 , 周园春 1, 2, *
计算机研究与发展 | 数据为中心的AI前沿进展及应用 2025,62(9): 2127-2138
收起
收藏切换
计算机研究与发展 |数据为中心的AI前沿进展及应用 2025 , 62 (9) : 2127 -2138
基于强化学习的科学数据特征生成算法
全屏
[Author(id=1327018485365977969, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=shaow@cnic.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327018485470835571, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, authorId=1327018485365977969, language=EN, stringName=Meng Xiao, firstName=Meng, middleName=null, lastName=Xiao, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1Computer Network Information Center, Chinese Academy of Sciences, Beijing 100190, bio={"content":"

born in 1995. PhD, assistant professor. Member of CCF. His main research interests include AI4Science and data-centric AI

"}, bioImg=null, bioContent=

born in 1995. PhD, assistant professor. Member of CCF. His main research interests include AI4Science and data-centric AI

, aboutCorrespAuthor=null), CN=AuthorExt(id=1327018485550527348, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, authorId=1327018485365977969, language=CN, stringName=肖濛, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1中国科学院计算机网络信息中心 北京 100190, bio={"img":"JoUPyfGRDgczEV1mKvjZ4Q==","content":"

1995年生. 博士,助理研究员. CCF会员. 主要研究方向为科学智能、以数据为核心的人工智能技术

"}, bioImg=JoUPyfGRDgczEV1mKvjZ4Q==, bioContent=

1995年生. 博士,助理研究员. CCF会员. 主要研究方向为科学智能、以数据为核心的人工智能技术

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327018483600175978, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, xref=1, ext=[AuthorCompanyExt(id=1327018483608564587, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018483600175978, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Computer Network Information Center, Chinese Academy of Sciences, Beijing 100190), AuthorCompanyExt(id=1327018483612758892, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018483600175978, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1中国科学院计算机网络信息中心 北京 100190)])]), Author(id=1327018485638607734, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327018485730882424, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, authorId=1327018485638607734, language=EN, stringName=Junfeng Zhou, firstName=Junfeng, middleName=null, lastName=Zhou, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2University of Chinese Academy of Sciences, Beijing 100049, bio={"content":"

born in 2002. Undergraduate. His main research interest includes data mining

"}, bioImg=null, bioContent=

born in 2002. Undergraduate. His main research interest includes data mining

, aboutCorrespAuthor=null), CN=AuthorExt(id=1327018485806379897, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, authorId=1327018485638607734, language=CN, stringName=周骏丰, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2中国科学院大学 北京 100049, bio={"img":"zQt8Elka1DFf4S5xRqMJ5w==","content":"

2002年生. 本科生. 主要研究方向为数据挖掘

"}, bioImg=zQt8Elka1DFf4S5xRqMJ5w==, bioContent=

2002年生. 本科生. 主要研究方向为数据挖掘

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327018485282091885, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, xref=2, ext=[AuthorCompanyExt(id=1327018485286286190, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018485282091885, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2University of Chinese Academy of Sciences, Beijing 100049), AuthorCompanyExt(id=1327018485303063407, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018485282091885, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2中国科学院大学 北京 100049)])]), Author(id=1327018485886071675, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=zyc@cnic.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1327018485986734974, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, authorId=1327018485886071675, language=EN, stringName=Yuanchun Zhou, firstName=Yuanchun, middleName=null, lastName=Zhou, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, *, address=1Computer Network Information Center, Chinese Academy of Sciences, Beijing 100190
2University of Chinese Academy of Sciences, Beijing 100049, bio={"content":"

born in 1975. PhD, professor, PhD supervisor. Senior member of CCF. His main research interests include scientific data and data intelligence

"}, bioImg=null, bioContent=

born in 1975. PhD, professor, PhD supervisor. Senior member of CCF. His main research interests include scientific data and data intelligence

, aboutCorrespAuthor=null), CN=AuthorExt(id=1327018486058038143, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, authorId=1327018485886071675, language=CN, stringName=周园春, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, *, address=1中国科学院计算机网络信息中心 北京 100190
2中国科学院大学 北京 100049, bio={"img":"bLukmfVLtoW5nY8TNPKQ+A==","content":"

1975年生. 博士,研究员,博士生导师. CCF高级会员. 主要研究方向为科学数据与数据智能

"}, bioImg=bLukmfVLtoW5nY8TNPKQ+A==, bioContent=

1975年生. 博士,研究员,博士生导师. CCF高级会员. 主要研究方向为科学数据与数据智能

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327018483600175978, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, xref=1, ext=[AuthorCompanyExt(id=1327018483608564587, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018483600175978, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Computer Network Information Center, Chinese Academy of Sciences, Beijing 100190), AuthorCompanyExt(id=1327018483612758892, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018483600175978, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1中国科学院计算机网络信息中心 北京 100190)]), AuthorCompany(id=1327018485282091885, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, xref=2, ext=[AuthorCompanyExt(id=1327018485286286190, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018485282091885, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2University of Chinese Academy of Sciences, Beijing 100049), AuthorCompanyExt(id=1327018485303063407, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327018481364611924, companyId=1327018485282091885, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2中国科学院大学 北京 100049)])])]
肖濛1 , 周骏丰2, 周园春1, 2, *
作者信息
  • 1中国科学院计算机网络信息中心 北京 100190
  • 2中国科学院大学 北京 100049
通讯作者:
周园春()
作者简介:

1995年生. 博士,助理研究员. CCF会员. 主要研究方向为科学智能、以数据为核心的人工智能技术

2002年生. 本科生. 主要研究方向为数据挖掘

1975年生. 博士,研究员,博士生导师. CCF高级会员. 主要研究方向为科学数据与数据智能

Reinforcement Learning-Based Feature Generation Algorithm for Scientific Data
Meng Xiao1 , Junfeng Zhou2, Yuanchun Zhou1, 2, *
Affiliations
  • 1Computer Network Information Center, Chinese Academy of Sciences, Beijing 100190
  • 2University of Chinese Academy of Sciences, Beijing 100049
  • born in 1995. PhD, assistant professor. Member of CCF. His main research interests include AI4Science and data-centric AI

    born in 2002. Undergraduate. His main research interest includes data mining

    born in 1975. PhD, professor, PhD supervisor. Senior member of CCF. His main research interests include scientific data and data intelligence

出版时间: 2025-09-01 doi: 10.7544/issn1000-1239.202550306
文章导航
收藏切换

特征生成(feature generation, FG)的目标是通过构建高阶特征组合并去除冗余特征,提升原始数据的预测潜力. 针对表格型科学数据,特征生成是提高下游机器学习模型性能的重要预处理环节. 然而,传统方法在处理科学数据特征生成问题时面临以下两方面挑战:首先,针对科学数据生成有效的高阶特征组合通常需要深入且广泛的领域知识;其次,随着特征组合阶数的增加,组合搜索空间呈指数级扩张,导致人工探索成本过高. 近年来,以数据为核心的人工智能(data-centric artificial intelligence, DCAI)范式的兴起为自动化特征生成过程提供了新的可能性. 受此启发,重新审视了传统特征生成工作流程,并提出了一种多智能体特征生成(multi-agent feature generation, MAFG)框架. 具体而言,在迭代探索阶段,多个智能体协作构建数学变换算式,识别并合成具有高信息含量的特征组合,最后通过强化学习机制实现策略的自适应演化. 探索阶段结束后,MAFG框架引入大语言模型(large language models, LLMs),针对探索过程中的每个关键模型性能突破点,解释性评估该步骤新生成的特征. 实验结果和具体案例研究表明,MAFG框架能够有效地实现特征生成过程的自动化,并能显著提升下游多个科学数据挖掘任务的效果.

深度强化学习  /  多智能体  /  以数据为核心的人工智能  /  特征生成  /  科学数据

Feature generation (FG) aims to enhance the prediction potential of original data by constructing high-order feature combinations and removing redundant features. It is a key preprocessing step for tabular scientific data to improve downstream machine-learning model performance. Traditional methods face the following two challenges when dealing with the feature generation of scientific data: First, the effective construction of high-order feature combinations in scientific data necessitates profound and extensive domain-specific expertise. Secondly, as the order of feature combinations increases, the search space expands exponentially, imposing prohibitive human labor consumption. Advancements in the data-centric artificial intelligence (DCAI) paradigm have opened novel avenues for automating feature generation processes. Inspired by that, this paper revisits the conventional feature generation workflow and proposes the multi-agent feature generation (MAFG) framework. Specifically, in the iterative exploration stage, multi-agents will construct mathematical transformation equations collaboratively, synthesize and identify feature combinations exhibiting high information content, and leverage a reinforcement learning mechanism to evolve their strategies. Upon completing the exploration phase, MAFG integrates the large language models (LLMs) to interpretatively evaluate the generated features of each significant model performance breakthrough. Experimental results and case studies consistently demonstrate that MAFG framework effectively automates the feature generation process and significantly enhances various downstream scientific data mining tasks.

deep reinforcement learning  /  multi-agent  /  data-centric AI  /  feature generation  /  scientific data
肖濛, 周骏丰, 周园春. 基于强化学习的科学数据特征生成算法. 计算机研究与发展, 2025 , 62 (9) : 2127 -2138 . DOI: 10.7544/issn1000-1239.202550306
Meng Xiao, Junfeng Zhou, Yuanchun Zhou. Reinforcement Learning-Based Feature Generation Algorithm for Scientific Data[J]. Journal of Computer Research and Development, 2025 , 62 (9) : 2127 -2138 . DOI: 10.7544/issn1000-1239.202550306
随着大数据的时代的到来,各个科学领域的数据集的规模不断增大,复杂性也不断上升. 如何有效地利用数据集中的特征信息成为了重要研究的问题. 特征生成是表格类科学数据(tabular scientific data)预处理中关键的一环,例如科学家通过先验知识与数学公式,组合身高、体重等特征,构建身体质量指数(body mass index,BMI)这一关键特征来诊断患糖尿病的风险. 然而对于大数据时代的科学数据集而言,传统的手工特征生成方法由于对专家经验的巨大依赖,以及受限于科学数据集复杂性而导致的巨大人力成本,导致其应用范围有限.
近年来,随着人工智能技术的进步,处理各类研究任务的机器学习模型逐渐成熟,研究视角从设计更为精巧的模型转移到以数据为核心的人工智能技术,并以设计更高质量的特征[1]为目标,这为自动化的特征生成带来了新的驱动力. 在这样的背景下,深度强化学习在单一智能体场景中取得了显著的成果[2],多智能体强化学习(multi-agent reinforcement learning,MARL)作为其扩展[3-4],逐渐被应用于多智能体协同决策问题. MARL允许多个智能体在1个系统中独立行动,并通过与环境交互和奖励信号来优化其策略. 将MARL应用于特征工程,可以将特征生成的决策过程建模为马尔可夫决策过程(Markov decision process, MDP),并利用多智能体的协同决策能力来利用数学变换组织特征,生成新的高效特征. 各智能体可以通过与环境交互学习到有效的特征变换策略,通过奖励机制优化智能体参数,提高下游模型机器学习的性能. 然而,基于强化学习范式的自动特征工程也有其天然的局限性. 这主要来源于其产出的高阶变换特征难以解释、在实际生产应用中难以落地.
基于以上讨论,本研究提出了一种基于多智能体强化学习的科学数据特征生成与自我解释框架,该框架通过引入强化学习模型,将特征与特征的组合过程视为一个多智能体的MDP. 3个智能体通过策略函数决定选择什么特征,以及通过一元或二元变换来组合、生成特征. 此外,本研究引入了大语言模型,对特征生成的关键性能突破点进行了可解释性评估,提升了生成特征的科学意义. 本工作的研究意义主要体现在3个方面,首先为科学数据集的特征工程提供了一个新的视角和方法,打破了传统方法在处理数据集时的局限性. 其次,通过实验和案例研究,验证了所提方法的有效性和鲁棒性,为相关科学领域的研究提供了可解释性评估. 最后,该方法能够在所选数据集上实现科学数据优化,并显著提升下游机器学习任务的效果,为科学数据的挖掘和分析提供了有力支持.
自动特征生成(automated feature generation,AFG)作为提升特征表达能力的有效手段,旨在通过数学运算、围绕数据本身,对原始特征进行空间扩展与重构,已成为机器学习预处理环节的重要研究方向[5]. 当前研究主要沿着下面3个技术路径展开:1)基于扩展-缩减的方法通过显式特征生成策略[6]或贪婪搜索算法[7]对原始特征空间进行组合扩展,继而采用特征选择机制进行空间约简,这类方法在多项式组合生成能力和复杂特征评估效率方面存在显著局限性;2)进化-评估方法将特征生成与选择过程整合为闭环学习系统,借助进化算法或强化学习模型进行迭代优化[8-10],但其离散决策机制导致计算复杂度高且收敛稳定性不足;3)基于自动机器学习(AutoML)的范式尝试将特征工程转化为自动化建模任务[2,11],虽然在特征构造自动化方面取得进展,但其特征生成阶次受限且变换质量呈现较大波动性,并由于时间复杂度过高,难以应用于大规模数据上. 此外,以上方法主要关注于生成过程,缺失了解释性评价,导致产出的高阶特征常常不具有人工可读性,优化后的数据集仅能供模型训练而无法直接应用于科学数据挖掘中最关注的知识发现. 总结而言,随着特征组合阶数的提升,特征空间呈指数级扩展,将导致搜索空间和计算复杂度迅速增加. 总结而言,随着特征组合阶数的提升,特征空间呈指数级扩展,导致搜索空间和计算复杂度迅速增加. 尽管多智能体机制和强化学习策略在一定程度上有助于缩小探索范围,但面对极大规模或高维科学数据,组合爆炸仍可能带来巨大的计算资源消耗和搜索效率下降,影响实际应用的可扩展性. 此外,现有方法对高阶变换特征的理论分析和数学有效性刻画尚不充分,缺乏系统性的解释性评价. 针对这些挑战,本文引入大语言模型(large language models,LLMs)作为知识驱动的解释与筛选机制,自动对生成特征进行可解释性评估,并在特征空间扩展过程中动态去除难以解释或冗余的高维特征,从而提升特征生成的科学有效性与实际可用性.
强化学习算法可以大致分为2类[1],一类是基于模型的强化学习,另一类是无模型强化学习. 对于基于模型的强化学习,顾名思义即依赖环境模型的强化学习算法. 它通过利用环境模型来降低对交互数据的依赖,从而加速训练并提高学习效率. 其典型算法有基于动态规划的(dynamic programming,DP)算法[12]、基于模型价值迭代的(model-based value iteration,MBVI)算法[13]以及基于模型策略迭代的(model-based policy iteration,MBPI)算法[14].
对于无模型强化学习及不依赖环境模型的强化学习算法,经典算法包括 Q-learning、SARSA(state-action-reward-state-action)等. Watkins等人[15]于1992年提出的Q-learning作为强化学习发展的基石对后面的算法起到了推动作用. 在Q-learning的基础上,2013年Mnih等人[16]提出了深度Q网络(deep Q-learning network,DQN),这些方法将深度神经网络引入Q-learning算法,并使用神经网络来估计Q值函数,通过这样的方式来对复杂状态空间进行学习和表示. DQN是一种将深度学习与Q学习相结合的算法,它能够在高维状态空间下进行决策. 其核心思想是使用神经网络来近似Q值函数,通过经验回放和目标网络的更新来提高学习的稳定性和效率. 该网络通过学习获得的奖励(reward)来对参数进行优化,以获得函数的最大Q值,从而使得智能体能够在动态的系统环境中学习和决策. 此外还有一些将2种思想融合的算法,如AC(actor-critic)算法、A2C(advantage actor-critic)算法、A3C(asynchronous actor-critic)算法[17-18]等.
特征生成旨在通过数学变换来组合特征以生成高阶特征,设基础特征集合$ F=\{{f}_{1},… ,{f}_{m}\} $,特征生成的数学操作符集合$ \mathcal{O}={\mathcal{O}}_{1}\cup {\mathcal{O}}_{2}\cup {\mathcal{O}}_{3} $由以下3类运算符构成. 下面分别对这3类运算符进行详细介绍.
1)一元运算符$ {\mathcal{O}}_{1} $. 对单个特征进行数学变换或预处理,$ {\mathcal{O}}_{1} $中包含的操作有$ \sqrt{x} $,$ {x}^{2} $,$ \mathrm{sin}\left(x\right) $,$ \mathrm{cos}\left(x\right) $,$ \mathrm{tanh}\left(x\right) $,$ \text{sigmoid}\left(x\right) $,$ \mathrm{log}\left(x\right) $,$ \dfrac{1}{x} $ .
2)二元运算符$ {\mathcal{O}}_{2} $. 对2个特征进行组合运算:$ {\mathcal{O}}_{2}=\{x+y, x-y, x\times y, x/y\} $,这些运算符可以将不同的特征进行组合,生成新的特征,从而挖掘特征之间的潜在关系.
3)预处理运算符$ {\mathcal{O}}_{3} $. 预处理运算符$ {\mathcal{O}}_{3}\subseteq {\mathcal{O}}_{1} $,包含标准化、归一化和分位数变换,这些方法是特征尺度变换操作的子集. 这些运算符支持在生成过程中动态调整数据分布,使得生成的特征更适合后续的机器学习任务. 其中标准化的定义为:$ {stand\_scaler}\left(x\right)= \dfrac{x-\mu }{\sigma } $,这里 $ \mu $ 是特征的均值,$ \sigma $ 是特征的标准差. 通过标准化操作,可以将特征的均值变为 0、标准差变为 1,使得不同特征具有相同的尺度,有助于机器学习模型的训练;归一化的定义为:$ {minmax\_scaler}\left(x\right)= \dfrac{x-\mathrm{min}\left(x\right)}{\mathrm{max}\left(x\right)-\mathrm{min}\left(x\right)} $,其目标在于将特征的值域映射到$ \left[\mathrm{0,1}\right] $区间. 归一化可以消除特征之间的量纲差异,使得所有特征在相同的尺度下进行比较. 分位数变换$ quan\_trans(x) $通过经验分布函数映射至均匀分布,这种变换可以使得特征的分布更加平滑,减少异常值对模型的影响.
特征变换序列是由一系列操作符和基础特征按照一定顺序组合而成的表达式. 在多智能体特征生成(multi-agent feature generation,MAFG)框架中,特征变换序列用于描述如何从基础特征集合 $ F $ 生成新的特征. 例如,对于基础特征 $ {f}_{1} $ 和 $ {f}_{2} $,一个可能的特征变换序列可以是 $ \tau ={(\sqrt{{f}_{1}}+{f}_{2})}^{2} $.
在多智能体强化学习中的状态表征用于描述当前环境的信息,而动作表征用于表示智能体可以采取的行动.
状态表征需要包含足够的信息,以便智能体能够做出合理的决策. 在MAFG框架中,状态包括当前的特征集合以及已经生成的特征变换序列.
动作表征表示智能体在当前状态下可以采取的行动. 在特征生成问题中,动作是选择某一个操作符或选择操作符所相应的特征进行组合(例如一元运算符需要选择1个特征,而二元操作符则需要选择2个特征),以生成新的特征变换序列. 例如,选择操作符“+”对特征 $ {f}_{1} $ 和 $ {f}_{2} $ 进行组合,就可以表示为一个动作.
通过操作符集合和操作符映射字典来实现动作的编码和解码. 智能体根据当前状态并通过相应的策略函数,生成新的特征变换序列.
科学数据特征生成框架的目标是找到最优的特征变换序列组合,从而提高下游机器学习模型的性能. 具体而言,给定原始数据集$ D={\left({x}_{i},{y}_{i}\right)}_{i=1}^{N} $,其中 $ {x}_{i} $ 表示样本特征,$ {y}_{i} $ 表示其标签或响应变量. 相关特征集合$ F $、下游机器学习模型 $ M $ 及其性能评价指标 $ V\left(\cdot \right) $(如准确率、AUC、MSE等),以及一组可选的特征变换操作集合 $ \mathcal{O} $,模型目标位自动地为特征集合 $ F $ 设计1条或多条特征变换序列$ \mathrm{\tau } $. 每个序列由不同的特征变换操作按照一定的顺序组成,作用于 $ F $ 上生成新的特征集合 $ F' $,并用于训练下游的机器学习模型,具体目标函数为
$ {\tau }^{*}=\mathop{\mathrm{arg\;max}}\limits_{\mathrm{\tau }}V\left(M\left(D,\mathrm{\tau }\left(F\right)\right)\right). $
如图1 所示,本研究提出的基于多智能体强化学习的科学数据特征生成框架MAFG通过3个协同工作的强化学习智能体实现动态特征生成与优化,其架构设计包含特征生成的探索、智能体的优化、关键特征组合的解释等模块. 本节将系统性阐述各模块的设计思想与交互机制.
MAFG框架以科学数据银行(science data bank)上的科学数据集为输入,并采用MinMaxScaler方法对数据进行归一化处理,对于输入特征矩阵$ {\boldsymbol{X}}\in {\mathbb{R}}^{n\times m} $($ n $为样本数,$ m $为特征维度),其归一化公式为:
$ \begin{aligned}{X}_{\text{norm}}^{\left(i,j\right)}=\;&\frac{{X}^{\left(i,j\right)}-\mathrm{min}\left({X}_{:,j}\right)}{\mathrm{max}\left({X}_{:,j}\right)-\mathrm{min}\left({X}_{:,j}\right)}\times \left(b-a\right)+a, \\&\forall i\in \{1,2,… ,n\},j\in \{1,2,… ,m\},\end{aligned} $
其中,$ a $和$ b $分别为目标区间的下限与上限,本框架设定$ a=-1 $, $ b=1 $. 该操作通过保留原始数据分布形态,将特征值映射至对称区间,有助于提升后续强化学习策略网络的训练稳定性. 归一化后的特征矩阵$ {{\boldsymbol{X}}}_{\text{norm}} $与目标变量$ y $重新组合为标准化数据集$ {\mathcal{D}}= \left[{{\boldsymbol{X}}}_{\text{norm}}|y\right] $,作为特征生成流程的输入,最终构建标准化数据表作为后续模块的输入.
一个基础的一元或者二元数学变换可以由2个必须的元素和1个候选元素组成. 在此基础上,MAFG框架设计了2个特征聚类智能体(Agent_C1, Agent_C2)与1个操作选择智能体(Agent_Op),三者通过状态-动作链式决策实现特征空间探索,以实现一元或者二元数学变换的生成,各智能体具体含义如下:
1)Agent_C1.基于当前特征聚类状态,从特征簇集合中选取初始特征子集.
2)Agent_Op.根据选定特征簇的嵌入表示,从预定义操作集$ \mathcal{O} $中选择特征变换算子.
3)Agent_C2.在二元操作场景下动态激活,基于操作语义与初始特征簇状态选择辅助特征簇,完成特征交叉计算.
智能体间采用参数隔离设计,通过差异化状态空间定义(如Agent_C2引入操作编码的扩展状态)实现决策解耦,同时共享全局奖励信号进行协同优化.
特征生成的核心在于分层强化决策与特征簇的动态组合,在特征聚类层面,采用无监督聚类算法将高维特征空间划分为语义相近的特征簇,聚类数目通过超参数调优确定. 该方法构建特征组的粗粒度表示,为强化学习提供可解释的决策单元.
在强化学习系统中,智能体的有效决策高度依赖于对环境状态的精准表征和关键特征的抽取. 因此,每个智能体的输入状态是当前数据集的统计量信息(例如标准差、均值、分位数等)与前步骤智能体的选择特征子集的统计量信息或所选择数学变换操作的独热编码(one-hot encoding).
在候选组合特征的选择层面,对于每回合(episode)中每一步(step)特征选择,系统采用阶段反馈驱动的策略进行参数更新. 首先由聚类智能体Agent_C1从特征聚类中选择一个聚类作为当前操作的特征子集,接着由操作选择智能体Agent_Op从操作集合中选择一个操作,如一元操作(${\cal O}_1 $)或二元操作(${\cal O}_2 $). 如果选择的是二元操作,另一个聚类智能体Agent_C2会从特征聚类中选择另一个聚类作为第2个操作数. 当生成特征维度超过预设阈值时,使用基于互信息统计量的特征选择算法进行特征重要性排序,保留对目标变量预测贡献度最高的特征子集,以及保留最重要的特征来有效控制特征空间爆炸问题. 3个智能体基于 DQN 系列算法[19-20]. DQN是一种将深度学习与Q-learning相结合的算法,它能够在高维状态空间下进行决策. 其核心思想是使用神经网络来近似Q值函数,通过经验回放和目标网络的更新来提高学习的稳定性和效率.
在DQN中,使用一个神经网络 $ Q\left(s,a;\mathrm{\theta }\right) $ 来估计状态 $ s $ 下执行动作 $ a $ 的Q值,其中 $ \mathrm{\theta } $ 是神经网络的参数. 训练的目标是最小化Q值估计与目标Q值之间的均方误差:
$ {L}\left(\mathrm{\theta }\right)={\mathrm{E}}_{\left(s,a,r,{s}'\right)\sim D}\left[{\left(r+\mathrm{\gamma }\mathrm{max}{a}'Q\left({s}',{a}';{\mathrm{\theta }}^-\right)-Q\big(s,a;\mathrm{\theta }\big)\right)}^{2}\right] \text{,} $
其中,$ D $ 是经验回放缓冲区,$ {\theta }^{-} $ 是目标网络的参数,$ \gamma $ 是折扣因子. 在下文中,我们还将介绍基于DQN算法的改进算法,并将该算法作为智能体强化学习方法上的策略优化方向.
最后,在特征组合层面,依据选定操作对特征簇实施符号计算,例如对2个特征簇进行逐元素相加生成高阶特征. 该过程引入特征命名规则化机制,保障生成特征的语义可追溯性. 在组合特征之后,如果特征的总数超过了阈值,则会对特征集合进行维度控制,采用特征选择算法智能地选择最优的特征子集.
强化学习奖励机制旨在通过下游预测任务的性能提升量,构建即时评估函数,将特征生成性能提升效果量化为强化学习信号. 对于分类类型的下游任务 (classification task),使用随机森林分类器,采用分层K折交叉验证,计算加权F1分数的平均值:
$ {P}_{\text{new}}^{\text{cls}}=\frac{1}{5}\sum_{k=1}^{5}{{F}{1}}_{k}^{\text{weig}\text{h}\text{ted}}\text{,} $
$ F1_k=2\times\frac{precision_k\times recall_k}{precision_k+recall_k}\text{,} $
其中,$ {{F}\text{1}}_{k}^{\text{weig}\text{h}\text{ted}} $ 表示第$ k $折交叉验证中按类别样本量加权的F1分数,计算覆盖全部类别.
对于回归类型的下游任务(regression task),框架使用随机森林回归器,计算相对绝对误差(1-RAE)的平均值,性能指标为:
$ {P}_{\text{new}}^{\text{reg}}=\frac{1}{5}\sum_{k=1}^{5}\left(1-{{RAE}}_{k}\right)\text{,} $
$ {{RAE}}_{k}=\dfrac{\displaystyle\sum\limits_{i\in {\mathcal{D}}_{{k}}}\left|{y}_{i}-\widehat{{y}}_{i}\right|}{\displaystyle\sum\limits_{i\in {\mathcal{D}}_{{k}}}\left|{y}_{i}-\overline{{y}}_{\text{train}}\right|} \text{,} $
其中,$ {\mathcal{D}}_{{k}} $为第$ k $折测试集,$ \overline{{y}}_{\text{train}} $为训练集目标变量均值,$ {{RAE}}_{k} $反映预测值$ \widehat{{y}}_{i} $相对于均值基准的相对绝对误差.
对于异常检测类型的下游任务(anomaly detection) ,框架将使用K近邻分类器,计算ROC AUC分数的平均值,性能指标如下:
$ {P}_{\text{new}}^{\text{det}}=\frac{1}{5}\sum_{k=1}^{5}{{AUC}}_{k}\text{,} $
$ AUC_k=\int_0^1TPR_k\left(FPR\right)\mathrm{d}FPR\text{,} $
其中,$ {{TPR}}_{k}\left({FPR}\right) $为第$ k $折的ROC曲线,$ {{AUC}}_{k} $表示模型区分正负类的能力.
对于以上3种任务,最终奖励函数可统一表达为:
$ {R}_{t}=\eta \times \left({P}_{\text{new}}-{P}_{\text{old}}\right) \text{,} $
其中,$ {P}_{\text{new}} $与$ {P}_{\text{old}} $按任务类型选择式(4)(6)(8)计算,$ \eta \in \left[\mathrm{0,1}\right] $为任务自适应的奖励缩放因子(默认$ \eta =1 $). 计算得到的奖励信号会通过分配给不同的动作或状态信号,并存储到相应的DQN的经验回放缓冲区中,用于后续的学习和更新.
各智能体会独立维护经验回放缓冲区和存储状态转移元组(状态,动作,奖励,下次状态). 当缓冲区达到容量阈值后,采用批量采样方式更新Q网络参数,通过目标网络延迟更新策略提升训练稳定性.
针对所生成的高阶特征评估的科学性和解释性问题,本文进一步设计并集成了基于大语言模型的特征评估模块. 如图1 右侧所示,MAFG系统在每一轮训练或决策阶段,会从记忆库(memory)中动态抽取最近的状态-动作-奖励历史序列,并结合奖励信息和决策动作,对相关特征进行评估.
为了定量指导特征生成过程,避免因智能体过拟合产生过于高阶组合特征的“幻觉”问题,系统将关键特征及其变换结果输入大语言模型,由模型输出关于特征重要性、相关性和作用机制的多维评价. 具体而言:1)对于每一组特征,框架将会通过奖励分数自动量化其对下游任务表现的影响力,避免仅凭大语言模型输出文本进行主观解读;2)通过与已有领域知识的交叉验证,特征评估模块对特征的合理性与科学性进行知识佐证,显著减少因框架过拟合生成幻觉解释的风险;3)框架将会去除过于复杂或是高奖励分但无法被大语言模型解释的特征变换,进一步提高特征评价的可靠性和可解释性.
综上,基于大语言模型的特征评估模块在强化学习系统中实现了对特征价值的定量判别和风险防控,有效避免了MAFG框架的解释幻觉问题,并为后续特征筛选、特征变换和决策优化提供了科学支持,提升了系统的泛化能力与整体解释性.
综上综述, MAFG框架的特征分组策略主要包含3个步骤:首先对特征进行聚类,将相似的特征划分到同一个组中;然后使用强化学习模型从聚类结果中选择合适的特征组;最后根据选择的操作对特征组进行处理,生成新的特征. 在产生关键的特征变换时,外置的大语言模型将会对这些关键变换进行可解释评价,从而使得变换过程中产生的知识能够被有效利用. 通过框架自动探索特征空间,选择最有价值的特征组进行特征生成,从而提高下游机器学习模型的性能.
本实验经过多维度调研,最终在科学数据银行1上挑选出3个实验对象数据集:气象与环境因子交互驱动的手足口病发病率数据集[21]、学习投入-师生关系协同驱动的学业表现数据集[22]、肾功能指标交互作用驱动的衰弱指数数据集[23]. 这些数据集均为包含连续性回归预测任务的表格类型数据(tabular data),且分别对应了公共卫生领域、教育领域和临床医学领域.
在模型训练的过程中,日志会记录下每一节中每一步的性能表现,并与之前的最优表现进行比较并更新最优表现. 在性能指标方面,本研究选取的回归评价指标主要有3个:平均绝对误差(mean absolute error,MAE)、均方根偏差(root mean squared error,RMSE)、相对绝对误差(1-relative absolute error,1-RAE).
MAE也称为 L1 损失,是损失函数的一种. 其计算方式是,先求出预测值与实际值之间的绝对差值,再对整个数据集的这些差值取平均值. MAE 衡量误差的大小,不考虑误差的方向. 一般来说,MAE越低,模型的准确性就越高. RMSE是通过取 MSE 的平方根来计算的,RMSE 越低,模型预测效果就越好. 1-RAE是总绝对误差与平均值和实际值之间的绝对差的比值. 1-RAE以比率的形式呈现,值的范围在0到1之间,并且越接近1越好.
表1中,在手足口病气象数据集的训练日志中,Episode 1-Step 3生成特征 temperature × precipitation 使模型1-RAE从原始数据的0.302跃升至0.358,提升幅度达 18.5%(其余2项指标提升不大),是全局性能最高跃升点.
表2中,在小学生学习投入数据集的训练日志中,Episode 2-Step 4生成特征ZT3dedication³ × ZT2TS²) / ZT3absorption使模型RMSE从0.013降至0.004,下降幅度为69.2%. 该特征在验证集的MAE从0.079降至0.038,误差下降幅度为 51.8%.
在胱抑素C与衰弱风险数据集的训练日志中,Episode 4-Step 4生成特征 (newcrp² + newhba1c) × cystatinc 使模型1-RAE从基准值 0.002跃升至0.035,提升幅度达16.5倍,是全局性能最高跃升点. 性能对比具体如表3所示.
总结而言,本实验通过对比MAFG框架生成特征的下游机器学习表现能力与原始数据的下游任务表现能力,论证了框架架构的有效性,在后续实验中,将会探讨强化学习算法的选择、框架中特征冗余去除模块,以及具体的性能跃升点的案例分析.
本实验通过改变框架智能体的强化学习算法做横向对比试验,并在3组不同类型的数据集上,通过相同轮次的探索与学习,对比了DQN、DDQN、DuelingDQN和DuelingDDQN四种强化学习方法在特征评价任务中的表现,并以原始数据的性能作为基线进行分析.
结果如图2所示. 首先可以发现各方法在不同数据集上的1-RAE(相对绝对误差)均显著高于原始数据,表明基于强化学习的特征评价方法能够有效提升模型性能. 经实验验证,模型性能如图2所示.
可见,不同强化学习算法在不同数据集上的优化效果不尽相同. 具体而言,在手足口病气象数据集上,DQN方法的1-RAE为0.358,优于原始数据的0.302,DDQN和DuelingDQN的性能略有提升,分别达到0.373和0.375,而DuelingDDQN取得了最高的1-RAE(0.392),提升最为显著. 在小学生学习投入数据集中,4种算法的表现进一步提升. DQN达到0.917,DDQN和DuelingDQN均为0.936,DuelingDDQN最高,为0.950,均高于原始数据的0.896,显示出强化学习方法对于复杂行为类数据的良好适应性. 在胱抑素C与衰弱风险数据集上,虽然1-RAE整体数值较小,但各算法同样优于原始数据(0.002). DQN、DDQN和DuelingDQN分别为0.009、0.010和0.010,DuelingDDQN提升至0.012,显示出强化学习方法在医学指标风险评估场景下的有效性.
总体而言,DuelingDDQN在相同轮次的探索和学习过程中,在所有数据集上均取得了最佳性能,验证了其在特征抽取与评价方面的优势. 同时,基于各种强化学习方法的MAFG框架均能显著提升模型的表现,进一步证明了该框架的有效性和通用性.
本实验以学习投入-师生关系协同驱动的学业表现数据集为实验对象,旨在系统评估所提出的MAFG框架在表格型教育数据中的特征冗余去除能力和高价值特征挖掘能力. 实验重点考察 MAFG 生成的新特征在不同特征选择方法辅助下,对提升下游学业表现预测模型的实际效果与特征利用率的贡献,从而验证以数据为中心的特征生成方法在科学数据建模中的实用价值.
在本实验中,MAFG分别结合 ExtraTrees、KBest、Lasso、RandomForest (RF)、Recursive Feature Elimination (RFE) 以及 None(不使用特征选择)这6种特征选择方式,构建对比实验组. 此外,为全面衡量新生成特征的实际利用率,设计了2类特征利用率评价指标.
1)平均特征利用率(average proportion,Proportion). 衡量每轮top-k特征中,新生成特征(即不属于原始特征集)的占比,定义为:
$ {Proportion}=\frac{\left|\{f\in T|f\notin F\}\right|}{\left|T\right|+\epsilon } \text{,} $
其中, $ T $ 为当前轮top-k特征集合,$ F $为原始特征集合,$ f $为一个特征,$ \epsilon $为极小常数. 越大的特征利用率Proportion表示本轮所生成的特征在 top-k 重要特征中的占比越大,代表生成特征越有效.
2)加权特征利用率(reward-weighted proportion,Weighted Proportion). 在平均特征利用率基础上,引入每轮性能提升信号,突出高性能场景下新特征的贡献,定义为:
$ {Weighted \;Proportion}={Proportion}\times \sigma \left(r\right) , $
其中,$ r $为当前轮奖励(reward),$ {\sigma }\left(r\right)=\dfrac{1}{1+{{\mathrm{e}}}^{-r}} $ 为Sigmoid函数. 越大的加权特征利用率表示本轮生成的特征不但能提升模型性能,同时也更为高效.
每项指标均取整个实验周期的平均值,以反映各特征选择方法下 MAFG 框架的特征冗余去除和高价值特征生成能力.
图3表明,MAFG 框架结合各类特征选择器均能提升下游1-RAE性能,并在合理的运行时间内完成特征生成. 尤其是在 ExtraTrees、Lasso 等方法下,特征生成的效果最为显著. 相比之下,None组(不使用特征选择模块)的运行时间明显更长,表明特征选择模块的去冗余功能能够有效加速特征生成过程. 对比None 组方法与其他特征选择组件的 1-RAE 分数,也充分展示了特征选择组件能够通过去除低质量的生成特征,提升最终生成数据集的品质. 在特征利用率方面,平均特征利用率与加权特征利用率均显示,MAFG生成的新特征在大多数特征选择器辅助下更容易被下游模型采纳,且高性能场景下新特征的实际贡献进一步提升. 这表明MAFG不仅能够去除冗余特征,还能有效挖掘和利用高价值特征.
总结而言,本实验系统验证了多智能体特征生成框架中特征选择组件的价值. 综合对比表明,MAFG框架联合特征选择策略,不仅能够提升模型预测性能,还能显著提升新特征的利用率和实际价值. 这一结果有力支持了以数据为中心的特征生成理念在实际应用场景中的有效性和普适性.
本节将围绕3组科学数据集进行结果讨论,结合模型定量的性能评估与特征评估模块,分析生成的特征组合在具体案例下的实际意义.
用KBest特征选择算法得出的数据集原特征重要性如表4所示.
经模型训练得到的训练日志反映出性能随步数增长所发生的变化趋势与关键性能变化步的特征生成组合,如图4所示. 其中标注出的关键性能变化步的特征生成组合如表5所示. 综合重要性和性能表现,选取重要性排名第一的特征组合temperature × precipitation,结合特征评估模块对该特征组合进行讨论. 讨论结果与生成特征可信来源如表6所示.
用特征选择模块得出的数据集原特征重要性如表7所示.
经模型训练得到的训练日志反映出性能随步数增长所发生的变化趋势,如图5所示. 其中标注出的关键性能变化步的特征生成组合如表8所示.
综合重要性和性能表现,选取重要性排名第2的特征组合(ZT3dedication3 × ZT2TS2) / ZT3absorption,结合特征评估模块对该特征组合进行讨论. 讨论结果与生成特征可信来源如表9所示. 从讨论结果来看,生成特征通过量化“奉献3×师生2/专注”的协同机制,突破传统模型的解释瓶颈,为建立“以学定教”的精准教育干预体系提供量化工具.
用特征选择模块得出的数据集原特征重要性如表10所示.
经模型训练得到的训练日志反映性能随步数增长所发生的变化趋势,如图6所示. 其中标注出的关键性能变化步的特征生成组合如表11所示. 综合重要性和性能表现,选取重要性排名第2的特征组合(newcrp² + newhbalc) × cystatinc,结合特征评估模块对该特征组合进行讨论. 讨论结果与生成特征可信来源如表12所示.
特征评估模块揭示了 (炎症2 + 代谢紊乱) × 肾功能损伤的系统效应,捕捉到CKD患者衰弱进展的核心病理三角,炎症与代谢紊乱的协同作用被肾功能衰减放大,形成恶性循环. 显著提升模型预测能力,并为医学临床的潜在精准干预提供新靶点.
本文针对大数据时代科学数据集特征复杂、人工特征生成方法局限性突出的现状,结合以数据为核心的人工智能视角,提出了一种基于多智能体强化学习与大语言模型相结合的自动特征生成与自我解释框架. 通过将特征生成过程建模为多智能体MDP,各智能体能够协同探索和优化特征组合策略,显著提升了特征工程的自动化和有效性. 引入大语言模型对关键特征及其变换过程进行评估与解释,不仅增强了模型的可解释性,也提升了生成特征的科学合理性和实用价值. 实验结果表明,所提方法在手足口病气象、小学生学习投入、胱抑素C与衰弱风险等多个科学数据集上均显著优于原始数据集,验证了本方法在特征生成和下游任务建模方面的有效性与鲁棒性. 此外,特征评估模块为特征的重要性分析和科学性验证提供了有力支持,推动了科学数据智能处理的进一步发展.
未来研究可进一步拓展多智能体强化学习在异构数据、多模态场景下的特征生成能力,并探索大语言模型与领域知识更加深度的融合,为科学数据挖掘和知识发现提供更为强大的智能工具.
作者贡献声明:肖濛提出了整体算法框架,进行了论文撰写;周骏丰完成了实验代码,参与了论文撰写;周园春进行了论文修改.

参考文献 引证文献
排序方式:
1
Zha Daochen, Bhat Z P, Lai K H, et al. Data-centric artificial intelligence: A survey[J]. ACM Computing Surveys, 2025, 57(5): 1−42
2
Wang Dongjie, Xiao Meng, Wu Min, et al. Reinforcement-enhanced autoregressive feature transformation: Gradient-steered search in continuous space for postfix expressions[C]// Proc of Conf on Neural Information Processing Systems. New Orleans, LA: NeurIPS, 2023: 43563−43578
3
李明阳, 许可儿, 宋志强, 等. 多智能体强化学习算法研究综述[J]. 计算机科学与探索, 2024, 18(8): 1979−1997
Li Mingyang, Xu Keer, Song Zhiqiang, et al. Review of research on multi-agent reinforcement learning algorithms[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(8): 1979−1997 (in Chinese)
4
陈人龙, 陈嘉礼, 李善琦, 等. 多智能体强化学习方法综述[J]. 信息对抗技术, 2024, 3(1): 18−32
Chen Renlong, Chen Jiali, Li Shanqi, et al. A survey of multi-agent reinforcement learning methods[J]. Information Countermeasure Technology, 2024, 3(1): 18−32 (in Chinese)
5
He Tianqi, Huang Xiaohan, Du Yi, et al. FastFT: Accelerating reinforced feature transformation via advanced exploration strategies[C]//Proc of 2025 IEEE 41st Int Conf on Data Engineering (ICDE). Los Alamitos, CA: IEEE Computer Society, 2025: 4120−4133
6
Katz G, Shin E C R, Song D. ExploreKit: Automatic feature generation and selection[C]//Proc of 2016 IEEE 16th Int Conf on Data Mining. Piscataway, NJ: IEEE, 2016: 979−984
7
Dor O, Reich Y. Strengthening learning algorithms by feature discovery[J]. Information Sciences, 2012, 189: 176−190
8
Xiao Meng, Wang Dongjie, Wu Min, et al. Traceable group-wise self-optimizing feature transformation learning: A dual optimization perspective[J]. ACM Transactions on Knowledge Discovery from Data, 2024, 18(4): 1−22
9
Khurana U, Samulowitz H, Turaga D. Feature engineering for predictive modeling using reinforcement learning[C/OL]//Proc of the AAAI Conf on Artificial Intelligence. Menlo Park, CA: AAAI, 2018[2025-07-09]. https://ojs.aaai.org/index.php/AAAI/article/view/11678/11537.
10
Xiao Meng, Wang Dongjie, Wu Min, et al. Traceable automatic feature transformation via cascading actor-critic agents[C]//Proc of the 2023 SIAM Int Conf on Data Mining (SDM). Philadelphia, PA: Society for Industrial and Applied Mathematics, 2023: 775−783
11
Zhu Guanghui, Xu Zhuoer, Yuan Chunfeng, et al. DIFER: Differentiable automated feature engineering[C]// Proc of the 1st Int Conf on Automated Machine Learning. New York: PMLR, 2022, 188: 17/1−17
12
Bellman R. Dynamic programming[J]. Science, 1966, 153(3731): 34−37
13
马琦, 刘杨, 吴贤生, 等. 基于价值迭代算法的最优渗透路径发现. 计算机系统应用, 2023, 32(12): 197−204
Ma Qi, Liu Yang, Wu Xiansheng, et al. Optimal penetration path discovery based on value iterative algorithm[J]. Computer Systems Applications, 2023, 32(12): 197−204 (in Chinese)
14
Ding Zihan, Huang Yanhua, Yuan Hang, et al. Introduction to reinforcement learning[J]. Deep Reinforcement Learning: Fundamentals, Research and Applications, 2020: 47−123
15
Watkins C J C H, Dayan P. Q-learning[J]. Machine Learning, 1992, 8: 279−292
16
Mnih V, Kavukcuoglu K, Silver D, et al. Playing Atari with deep reinforcement learning[J]. arXiv preprint, arXiv: 1312.5602, 2013
17
Sutton R S, McAllester D, Singh S, et al. Policy gradient methods for reinforcement learning with function approximation[C/OL]// Proc of Conf on Neural Information Processing Systems. Denver, CO, USA: NeurIPS, 1999[2025-07-09]. https://proceedings.neurips.cc/paper/1713-policy-gradient-methods-for-reinforcement-learning-with-function-approximat.
18
Mnih V, Badia A P, Mirza M, et al. Asynchronous methods for deep reinforcement learning[C]//Proc of the 33rd Int Conf on Machine Learning. New York: PMLR, 2016: 1928−1937
19
况立群, 冯利, 韩燮, 等. 基于双深度网络的智能决策系统研究[J]. 计算机技术与发展, 2022, 32(2): 137−142
Kuang Liqun, Feng Li, Han Xie, et al. Research on intelligent decision-making system based on double deep Q-Network[J]. Computer Technology and Development, 2022, 32(2): 137−142 (in Chinese)
20
张俊杰, 张聪, 赵涵捷. 重复利用状态值的竞争深度网络算法[J]. 计算机工程与应用, 2021, 57(4): 134−140
Zhang Junjie, Zhang Cong, Zhao Hanjie. Dueling deep Q network algorithm with state value reuse[J]. Computer Engineering and Applications, 2021, 57(4): 134−140 (in Chinese)
21
Liu Yunxia. Data set of hand, foot and mouth disease and main meteorological factors at county level in Shandong Province, China, 2015−2019[DS/OL]. V1. Science Data Bank, 2023[2025-07-09]. https://cstr.cn/31253.11.sciencedb.07251
22
Yu Xiao. Data set of learning engagement and environmental variables of Chinese Primary School children[DS/OL]. V2. Science Data Bank, 2023[2025-07-09]. https://cstr.cn/31253.11.sciencedb.06343
23
Xing Yuping, Xing Hui, Li Miao, et al. Relationship between cystatin C levels and risk of frailty in two Chinese populations aged 45~65 years and 65 years or older[DS/OL]. V1. Science Data Bank, 2023[2025-07-09]. https://cstr.cn/31253.11.sciencedb.o00130.00340
24
Tsou Y L, Lin Y W, Shao H Y, et al. Recombinant adeno-vaccine expressing enterovirus 71-like particles against hand, foot, and mouth disease[J]. PLoS Neglected Tropical Diseases, 2015, 9(4): e0003692
25
Nguyen A T, Grembi J A, Riviere M, et al. Influence of temperature and precipitation on the effectiveness of water, sanitation, and handwashing interventions against childhood diarrheal disease in rural Bangladesh: A reanalysis of the WASH Benefits Bangladesh trial[J]. Environmental Health Perspectives, 2024, 132(4): 047006
26
Andhikaputra G, Sapkota A, Lin Y K, et al. The impact of temperature and precipitation on all-infectious-, bacterial-, and viral-diarrheal disease in Taiwan[J]. Science of The Total Environment, 2023, 862: 160850
27
Wesenberg L, Jansen S, Krieglstein F, et al. The influence of seductive details in learning environments with low and high extrinsic motivation[J]. Learning and Instruction, 2025, 96: 102054
28
Garrote A, Diener M, Hepberger B, et al. Social behavior, academic achievement, language skills, and peer rejection in elementary school classes: The moderating role of teacher feedback[J]. Educational Psychology, 2024, 44(5): 613−631
29
Kong Yangtao. The role of experiential learning on students’ motivation and classroom engagement[J]. Frontiers in Psychology, 2021, 12: 771272
30
Bukabau J B, Yayo E, Gnionsahé A, et al. Performance of creatinine-or cystatin C–based equations to estimate glomerular filtration rate in sub-Saharan African populations[J]. Kidney International, 2019, 95(5): 1181−1189
31
Rudberg S, Rasmussen L M, Bangstad H J, et al. Influence of insertiondeletion polymorphism in the ACE-I gene on the progression of diabetic glomerulopathy in type 1 diabetic patients with microalbuminuria[J]. Diabetes Care, 2000, 23(4): 544−548
32
Lim S K, Kim S M I, Lee E K, et al. Usefulness of cystatin C in measuring renal function in patients with liver cirrhosis[J]. Nephrology Dialysis Transplantation, 2024, 39(Supplement_1): g−f
2025年第62卷第9期
108
46
0
引用本文
BibTeX
文章信息
doi: 10.7544/issn1000-1239.202550306
  • 接收时间:2025-05-07
  • 首发时间:2026-11-10
  • 出版时间:2025-09-01
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2025-05-07
  • 修回日期:2025-07-09
基金
作者信息
    1中国科学院计算机网络信息中心 北京 100190
    2中国科学院大学 北京 100049

通讯作者:

周园春()
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202550306
分享至
全文二维码

扫描看全文

引用本文
BibTeX
本文的引用情况
2种不同金属材料的力学参数
科
Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)
属
Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏