基于弱关系的异质社交网络推荐  

Heterogeneous social network recommendation based on weak ties

作　　者：荀亚玲[1] 毕慧敏 张继福[1] XUN Ya-ling;BI Hui-min;ZHANG Ji-fu(School of Computer Science and Technology,Taiyuan University of Science and Technology,Taiyuan 030024,China)

机构地区：[1]太原科技大学计算机科学与技术学院,山西太原030024

出　　处：《计算机工程与设计》2021年第6期1526-1534,共9页Computer Engineering and Design

基　　金：国家青年科学基金项目(61602335);山西省自然科学基金项目(201901D211302);太原科技大学博士科研启动基金项目(20172017)。

摘　　要：针对基于“相似性”或深度信任的传统强关系推荐导致的推荐结果趋于同质性问题,利用弱关系丰富的语义信息和强大的信息传递能力,提出一种基于弱关系的异质社交网络推荐算法。借助用户间的信任值对因引入弱关系产生的大量用户节点进行筛选,减少后续计算负载;充分利用网络中丰富的对象和关系信息,建立拓展的全关系用户-项目异质信息网络模型(UI-HIN),改善推荐的单一性与冗余性;利用最佳信任路径选择算法(BTP)对全关系UI-HIN路径搜索进行优化得到一组最佳信任路径。实验结果表明,该算法的推荐效果具有更好的针对性、新颖性和多样性。Aiming at the problem of homogeneity of recommendation results caused by traditional strong relationship recommendation based on similarity or deep trust,a recommendation algorithm based on weak ties was proposed using rich semantic information and powerful information transmission capabilities of weak ties.The trust values between users were used to filter trust nodes to reduce the subsequent calculation workload.An extended full-relational user-item heterogeneous information network model(UI-HIN)was constructed based on the rich object and relationship information in the network,which effectively improved the unity and redundancy of recommendations.The best trust path selection algorithm(BTP)was proposed to optimize the path search of the full-relationship UI-HIN to obtain a set of optimal trust paths.Experimental results show the recommendation algorithm has better performance in pertinence,novelty and diversity.

关 键 词：弱关系 推荐算法 社交网络 异质网络 最佳信任路径 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]