摘要
[目的/意义]为了解决由于好友推荐缺乏对推荐结果解释导致用户无法评估推荐好友的质量,进而影响他们做出更好决策的问题。[方法/过程]首先利用LDA模型提取用户博文主题,计算余弦相似度得到博文相似度;通过用户共同好友比例计算好友相似度;运用Jaccard算法计算研究领域相似度。然后将以上三种相似度融合以计算用户相似度,并设计了基于博文主题、共同好友和研究领域的文本解释,最后融合用户相似度与文本解释,在提供好友推荐列表的同时提供文本解释。[结果/结论]模型不仅提高了好友推荐的准确性,而且通过提供解释帮助用户做出更好的决策,从而提高好友推荐的质量和用户满意度。[创新/局限]本研究的创新之处在于将可解释性引入到好友推荐领域,增强了用户对推荐结果的理解和接受度,从而做出更好的决策。但未考虑文本解释长度对解释有效性的影响,将在后续研究中进一步探讨。
[Purpose/significance]To address the issue where the lack of explanations for friend recommendation results hinders users from assessing the quality of the recommended friends,thereby impacting their ability to make better decisions.[Method/process]The process begins with the extraction of user blog topics using the LDA model,followed by the calculation of blog content similarity through cosine similarity;friend similarity is determined based on the proportion of mutual friends;and research field similarity is com-puted using the Jaccard algorithm.These three types of similarities are then integrated to calculate an overall user similarity.Textual explanations based on blog topics,mutual friends,and research fields are subsequently designed.Finally,user similarity and textual explanations are merged to provide textual explanations alongside the friend recommendation list.[Results/conclusion]The model not only enhances the accuracy of friend recommendations but also aids users in making more informed decisions by providing explana-tions,thus improving the quality of friend recommendations and increasing user satisfaction.[Innovation/limitations]The innovation of this study lies in the introduction of explainability into the realm of friend recommendation,which enhances users'understanding and acceptance of the recommendation results,enabling them to make better decisions.However,the impact of the length of textual ex-planations on their effectiveness was not considered,which will be further explored in future research.
作者
杨瑞仙
刘莉莉
于政杰
金燕
YANG Ruixian;LIU Lili;YU Zhengjie;JIN Yan(School of Information Management,Zhengzhou University,Zhengzhou 450001,China;Research Institute of Data Science,Zhengzhou city,Zhengzhou 450001,China;Central Big Data Innovation Center,China Academy of Information and Communication,Zhengzhou 450001,China)
出处
《情报科学》
CSSCI
北大核心
2024年第7期84-96,共13页
Information Science
基金
教育部哲学社会科学研究后期资助项目“融入多用户属性的网络知识社区核心用户识别与推荐研究”(23JHQ085)
河南省高等学校哲学社会科学创新团队支持计划“数据治理与交易流通”(2024-CXTD-01)
河南省高等学校青年骨干教师培养计划项目“学术虚拟社区知识交流机制的系统动力学仿真研究”(2020GGJS012)。
关键词
好友推荐
可解释推荐系统
LDA模型
相似度计算
解释的有效性
friend recommendation
explainable recommendation systems
LDA model
similarity calculation
explanation effective-ness
作者简介
杨瑞仙(1982-),女,河南济源人,博士,教授,博士生导师,主要从事信息计量与科学评价、社会网络与知识交流、数据隐私与数据治理研究;刘莉莉(1995-),女,青海海东人,硕士研究生,主要从事社会网络与社会计算研究;于政杰(1990-),男,河南郑州人,硕士研究生,主要从事社会网络与社会计算研究;金燕(1977-),女,河南周口人,博士,教授,博士生导师,主要从事数据治理、用户信息行为研究。
