社会化标注系统中个性化信息推荐多维度融合与优化模型研究  

Research on the Multi-dimensional Fusion and Optimization Model of Personalized Information Recommendation in Social Tagging System

作　　者：武慧娟[1] 孙鸿飞[1,2] 金永昌 Wu Huijuan;Sun Hongfei;Jin Yongchang(School of Economics and Management, Northeast Dianli University, Jilin 132012, China;School of Management, Jilin University, Changchun 130022, China)

机构地区：[1]东北电力大学经济管理学院,吉林吉林132012 [2]吉林大学管理学院,吉林长春130022

出　　处：《现代情报》2019年第1期37-42,85,共7页Journal of Modern Information

基　　金：教育部人文社会科学规划项目"社会化标注系统中个性化信息推荐多维度融合与优化模型研究"(项目编号:15YJC870024);吉林省教育厅"十三五"社会科学研究规划项目"基于用户认知的多源融合个性化微阅读用户画像模型研究"(项目编号:JJKH20190715SK)

摘　　要：[目的/意义]在社会化标注系统自组织运行的基础上,构建个性化信息推荐的多维度融合与优化模型,进而在大数据环境下,为用户提供精准的个性化信息推荐服务,从而进一步丰富个性化信息推荐的理论体系以及拓展个性化信息推荐的研究方法。[方法/过程]首先,对每一种个性化信息推荐方法的优点和不足进行深入分析;然后,将基于图论(社会网络关系)、基于协同过滤以及基于内容(主题) 3种个性化信息推荐方法进行多维度深度融合,构建个性化信息推荐多维度融合模型;最后,对社会化标注系统中个性化信息推荐多维度融合模型进行优化,从而解决个性化推荐过程中用户"冷启动"、数据稀疏性和用户偏好漂移等问题。[结果/结论]通过综合考虑现有的基于图论(社会网络关系)、基于协同过滤以及基于内容(主题)的个性化信息推荐方法各自的贡献和不足,实现3种方法之间的多维度深度融合,并结合心理认知、用户情境以及时间、空间等优化因素,最终构建出社会化标注系统中个性化信息推荐多维度融合与优化模型。[Purpose/Significance]On the basis of the self-organizing operation of the social tagging system,the multi-dimensional fusion and optimization model of personalized information recommendation was constructed to provide accurate personalized information recommendation service for users under the big data environment,thus further enriching the theoretical system of personalized information recommendation and expanding the research methods of personalized information recommendation.[Method/Process]Firstly,the advantages and disadvantages of each personalized information recommendation method were analyzed in depth.Then,three personalized information recommendation methods based on graph theory(social network relationship),collaborative filtering and content-based(topic-based)were fused in depth to construct a personalized information recommendation multi-dimensional fusion model.Finally,the multi-dimensional fusion model of personalized information recommendation in social annotation system was optimized to solve the problems of cold start,data sparsity and users preference drift in personalized recommendation process.[Result/Conclusion]By considering the respective contributions and shortcomings of graph theory(social network relationship),collaborative filtering and content-based(topic-based)personalized information recommendation methods,the multi-dimensional and deep fusion among the three methods was realized.Finally,a multi-dimensional fusion and optimization model of personalized information recommendation in social annotation system was constructed by combining psychological cognition,user context,time and space.

关 键 词：个性化信息推荐 社会化标注 多维度融合 优化模型 

分 类 号：G203[文化科学—传播学] G252