基于用户偏好的向量化语义社区发现与合并  

Vectorized Semantic Community Detection and Merging Based on User Preferences

作　　者：金结良 孙美丽 杨汕 JIN Jieliang;SUN Meili;YANG Shan(Modern Business College,Anqing Vocational and Technical College,Anqing 246003,China;College of Information Technology and Artificial Intelligence,Zhejiang University of Finance and Economics,Hangzhou 310000,China)

机构地区：[1]安庆职业技术学院现代商务学院,安徽安庆246003 [2]浙江财经大学信息技术与人工智能学院,杭州310000

出　　处：《安阳工学院学报》2024年第6期58-64,共7页Journal of Anyang Institute of Technology

基　　金：安徽高校人文社会科学研究重点项目(SK2021A1013);安徽省省级质量工程(2021xnfzxm066);安徽高校人文社会科学研究项目(2022AH052594)。

摘　　要：社区发现已成为个性化推荐中必不可少的重要研究领域。然而,现有方法忽略了用户偏好与社区结构的有效结合,以及社区冗余带来的推荐压力。为此,提出一种新的基于用户偏好的向量化语义社区发现与合并的方法。首先,构建偏好知识库用于捕捉用户偏好。然后,基于偏好知识库和Leader Rank算法定位种子节点。最后,为研究合并前后社区共有主题性和紧密性的变化趋势,基于向量化主题和关键词合并高冗余度的社区。通过对参数实验、对比实验等实验结果的分析,该方法优于主流社区发现方法,有利于提高基于社区发现的推荐准确性和用户体验感。Community detection has become a crucial research area in personalized recommendation. However,existing methods often overlook the effective integration of user preferences with community structures, as well asthe recommendation bias introduced by community redundancy. To address these issues, we propose a method forvectorized semantic community detection and merging based on user preferences. First, a preference knowledgebase is constructed to capture user preferences. Second, seed nodes are identified using the preference knowledgebase and the LeaderRank algorithm. Finally, to examine he changing trends in the shared themes and compactnessof communities before and after merging, communities with high redundancy are merged based on vectorizedthemes and keywords. Through parameter and comparative experiments, we demonstrate that the proposed methodoutperforms mainstream community detection techniques, which can enhance the accuracy of community-basedrecommendations and improve the user experience.

关 键 词：社区发现 个性化推荐 社区冗余 用户偏好 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]