论文部分内容阅读
随着互联网和在线社交平台的飞速发展,社会网络开始发展壮大起来,针对社会网络的研究也变得越来越热门。特别是在大数据时代,社会网络研究的价值和意义也将越来越大。个人网络是一种特殊的社会网络,针对个人网络的社交圈发现算法研究意义重大。各种社交平台基本都允许用户将其朋友手动划分到不同的社交圈中,但是对个人网络社交圈进行自动划分的方法研究十分稀少,不断变庞大、变复杂的个人网络致使对其社交圈发现算法研究的难度也在不断提高。Julian等人于2012年首先提出了基于概率模型的个人网络社交圈发现算法(DSCEN算法),实验证明该算法可以有效发现个人网络中的社交圈,并允许社交圈存在重叠、嵌套。本文在研究DSCEN算法的基础上,加入节点相似度因素对算法进行改进,并同时考虑属性相似度和拓扑相似度保证节点相似度的准确性。通过实验验证,改进后的算法具有更好的效果。
With the rapid development of the Internet and online social platforms, social networks have started to grow and develop, and research on social networks has become more and more popular. Especially in the big data era, the value and significance of social network research will also be greater and greater. Personal network is a kind of special social network, which is of great significance to the social circle discovery algorithm of personal network. Various social platforms basically allow users to manually divide their friends into different social circles, but the method of automatically demarcating social networks of personal networks is scarce, and the ever-changing, ever-changing and complex personal network has led to the discovery of their social circles The difficulty of algorithm research is also increasing. Julian et al first proposed a personal network social circle discovery algorithm (DSCEN algorithm) based on probabilistic model in 2012, and experiments show that the algorithm can effectively find social circles in personal networks and allow social circles to overlap and nest. Based on the study of DSCEN algorithm, this paper improves the algorithm by adding the similarity of nodes and considers the similarity of attributes and topological similarity to ensure the accuracy of node similarity. Through experimental verification, the improved algorithm has better effect.