T-BOI：一种融合时间和行为顺序信息的序列推荐系统

打开文本图片集

中图分类号：TN711-34；TP391.41 文献标识码：A 文章编号：1004-373X（2025）13-0088-08

T-BOI： A sequence recommendation system integrating temporal andbehavioralorderinformation

ZHANG Jing，ZHOU Diao

（CollegeofArtificialIntellgence，NorthChina UniversityofScienceandTechnology，TangshanO63210，China）

Abstract：Recommendation systems basedonuser behavior modelingare widelyused inrecall，sorting，and otherstages， includingsequentialrecommendationsandsesionrecommendations.Inthesequencerecommendation，somebehaviorrecords thatarenotrelatedtothenextbehaviormaybeintroducedduetoexcesivelylongsequencesetingsHowever，thesesion-based recommendation focusesonmediumandshort-termrecommendations，soithaslimitationsincapturinglong-termorgeneral interests.Inviewoftheabove，asequencerecommendationsystem（T-BOl） thatintegratestemporalandbehavioralorder informationhasbeenproposed，makingitsuitableforlong-termandshort-terminterestrecommendations.Intheproposed method，thefeaturerepresentationunitmodule，behaviorweightunitmodule，behaviorsequencerepresentationmodel，and behaviorcategoryoutputunit modulein T-BOIareused to process andobtainthe final predictionresult.The proposed methodis comparedwithsomeadvancedmodels.Thevalidationonpublicdatasetsshowsthat therecommendationsystem hasgood recommendation effect.

Keywords：sequencerecommendation；recommendationsystem；multi behaviorrecommendation;longand short-term preference;position encoding;time information

0 引言

在网络高速发展的时代，互联网用户在网上留下了大量的历史行为信息，从这些信息中，往往希望能够分析出用户的个性偏好，以协助提供更佳的互联网服务。（剩余15189字）