基于兴趣的社交网络用户聚类及可视化

发布时间：2018-02-14 09:00

本文关键词： 社交网络聚类数据可视化潜在语义模型　出处：《计算机科学》2017年S2期 　论文类型：期刊论文

【摘要】：随着社交网络的流行,从各种各样的社交网络数据中提取出有效信息并进行清晰直观的可视化分析,从而为用户提供有价值的潜在知识,显得尤为重要。聚类分析是数据挖掘中的重要分析手段,传统的面向社交网络数据的用户聚类分析大都仅考虑网络的拓扑链接结构,未考虑用户的兴趣相似度。文中基于贝叶斯概率模型来计算用户兴趣相似度并进行聚类,进一步设计交互可视化方式来展示上述聚类结果。具体地,针对社交网络中的用户评分数据建立潜在语义模型来提取表示每个用户兴趣特点的特征向量;基于用户的特征向量对用户进行聚类,得到具有不同特征的人群,并通过实验和热度图选择合适的人群聚类数;最后提出了基于层次气泡图的可视化展现和分析方案,将用户、电影类型、电影等多维信息在图形中交互展示,支持用户从全局概览到局部细节的推进式探索,从多角度可视化人群特征。对豆瓣网用户和电影评分数据进行了实验和分析,结果验证了所提方法的有效性。
[Abstract]:With the popularity of social networks, effective information is extracted from a wide variety of social network data and analyzed visually and clearly, thus providing users with valuable potential knowledge. Clustering analysis is an important analysis method in data mining. The traditional user clustering analysis for social network data only considers the topology link structure of the network. The interest similarity of users is not considered in this paper. Based on Bayesian probabilistic model, the interest similarity of users is calculated and clustered, and further interactive visualization is designed to display the above clustering results. The latent semantic model is established to extract the feature vectors representing the characteristics of each user's interest, and the users are clustered based on the user's feature vectors to get the people with different characteristics. Through experiments and heat map to select the appropriate number of people clustering. Finally, a visual display and analysis scheme based on hierarchical bubble graph is proposed. The multi-dimensional information such as user, film type, film and other multi-dimensional information are interactively displayed in the graph. It supports users to explore from global overview to local details, and visualizes crowd characteristics from various angles. The experimental and analysis of the user and movie scoring data of Douban.com are carried out, and the results show that the proposed method is effective.
【作者单位】：浙江工业大学计算机科学与技术学院;南通大学理学院;浙江大学管理学院;
【基金】：国家教育部新世纪优秀人才支持计划(NCET-13-0526) 国家自然科学基金(71571160) 浙江省自然科学基金(LY14F020021)资助
【分类号】：TP311.13;TP393.09

【相似文献】