一种基于二部图和节点角色划分的社交网络推荐方案

针对现有社交网络用户推荐方案中大规模网络个体相似性计算复杂度高以及个体节点无差异对待的问题,本文提出一种基于二部图和节点角色划分的推荐方案.首先,通过划分重叠群体简化原生社交网络结构,并进一步构建群体-个体二部图模型;其次,通过群体-个体二部图所反映的拓扑特征,结合节点自身属性特征,对个体进行角色划分,提出一种基于群体-个体二部图的角色划分模型;最后,针对大规模网络中计算个体相似性复杂度高的问题,构建基于角色差异下的个体-个体二部图模型,实现层次化、个性化的推荐.实验表明,该方案适用于对社交网络中兴趣广泛度存在差异的个体间进行好友推荐,并在较小规模的二部图上生成目标个体推荐列表,降低了计算个体相似性的复杂度.

A Social Network Recommendation Scheme Based on Bipartite Graph and Node Role Division

In the view of the high complexity about similarity calculation and the indifference about individual nodes,a social network recommendation scheme based on bipartite graph and node role division is presented in this study.Firstly,the native social network structure is simplified by dividing overlapping groups.Furthermore,the bipartite graph model of group and individual is given.Secondly,the role division model is proposed by combining topological features of bipartite graph with node attributes.Finally,in order to resolve high computational complexity,the individual bipartite graph model is constructed based on user role difference.The model implements a hierarchical and personalized recommendation.Experiments show that the scheme can effectively recommend among social users who have different interests.In addition,the complexity of individual similarity computation is reduced because the target individual recommendation list is generated based on small scale bipartite graph.