摘要
尽管用户可自主生成个性化数据以更全面描述个人偏好,但由于用户创建数据不严谨、不可控,导致生成的庞大数据集大多存在质量低、噪声严重的缺陷.因此管理复杂网络信息时,不能仅使用写入性知识,必须重视具有大量领域知识的专家,因为其可为系统提供高质量的信息.本文通过构建和分析用户兴趣分布曲线以发现兴趣领域专家,并提出甄别状态不正常的伪专家算法.由于网络中权威专家数量较少,所以所提供的信息是有限的.因此本文定义的领域专家不仅包含权威专家,而且包含普通用户中对某领域有极高关注的兴趣领域专家.实验证明算法的正确性和高效性,并且较低的复杂度使其可处理海量用户节点信息.
Although users can self-generate personalized data for describing preferences more comprehensively, user-created data is not rigorous and uncontrollable, which leads to enormous data of low quality with serious noise. On managing complex network, more attentions should be placed on high quality information that has been or will be produced by experts who have knowledge in specific fields in order to avoid being restricted to written knowledge. This paper finds experts by constructing and analyzing interest profiles of users and proposes a screening method for detecting abnormal pseudo-experts. Due to the small number of authoritative experts in networks, which provide a limited amount of information,experts defined in this paper not only include authoritative experts, but also ordinary users that have a lot of knowledge in a certain field. Experiments illustrate the correctness and effectiveness of the algorithm,and the low complexity renders it suitable in handling massive user node information.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2015年第8期1561-1567,共7页
Acta Electronica Sinica
基金
国家自然科学基金(No.60973040)
国家自然科学青年基金(No.61300148)
吉林省重点科技攻关项目基金(No.20130206051GX)
吉林省科技计划青年科研基金(No.20130522112JH)
中国博士后基金项目(No.2012M510879)
吉林大学基本科研业务费科学前沿与交叉项目(No.201103129)
关键词
专家发现
兴趣分析
兴趣图谱
复杂网络分析
expert finding
interest analysis
interest graph
complex network analysis