融合多维特征与兴趣漂移的虚拟学术社区群推荐模型  

Virtual Academic Community Group Recommendation Model Integrating Multidimensional Features and Interest Drift

作　　者：魏玲[1] 权晨雪 Wei Ling;Quan Chenxue(School of Economics and Management,Harbin University of Science and Technology,Harbin 150040,China)

机构地区：[1]哈尔滨理工大学经济与管理学院,黑龙江哈尔滨150040

出　　处：《现代情报》2023年第7期48-63,共16页Journal of Modern Information

基　　金：黑龙江省自然科学基金项目“融媒体时代突发事件网络舆情引导机制研究”(项目编号:LH2019G017)。

摘　　要：[目的/意义]为促进虚拟学术社区知识流转效率,弥补当前针对核心用户群组识别及动态兴趣漂移关注不足的问题,本文构建融合多维特征与兴趣漂移的虚拟学术社区群推荐模型。[方法/过程]以“科学网”为研究对象,从网络传播维度和网络结构维度出发,运用改进的信息熵公式综合识别核心用户并聚类发现用户群组。在此基础上,融合群组兴趣特征向量,基于滑动时间窗口和非线性遗忘曲线分析群组兴趣漂移过程,根据动态兴趣漂移结果进行群组推荐并验证该模型的适用性。[结果/结论]实验结果表明,该模型基于用户多维特征可以准确识别核心用户并能很好地反映群组兴趣漂移特征,同时,本文提出的群组推荐算法相比传统算法在推荐结果的准确率上明显提升。[Purpose/Significance]In order to promote the efficiency of knowledge transfer in virtual academic community and make up for the current problem of insufficient attention to core user identification and its dynamic interest drift,this paper constructs a recommendation model of virtual academic community group that integrates multidimensional characteristics and interest drift.[Method/Process]Taking“ScienceNet”as the research object,starting from the network propagation dimension and network structure dimension,the improved information entropy formula was used to comprehensively identify core users and cluster to find user groups.On this basis,the group interest feature vector was fused,the group interest drift process was analyzed based on the sliding time window and nonlinear forgetting curve,and the group recommendation was made according to the dynamic interest drift results and the applicability of the model was verified.[Result/Conclusion]The experimental results show that the model can accurately identify the core users in the community based on the multidimensional characteristics of users and well reflect the characteristics of group interest drift.At the same time,the group recommendation algorithm has significantly improved the accuracy of recommendation results compared with the traditional algorithm.

关 键 词：虚拟学术社区 核心用户 偏好融合 兴趣漂移 群推荐 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]