多关系网络上的社区发现

Community detection in multi⁃relational networks

真实的网络结构上的节点之间通常存在多种类型的边,依据网络上的各种边对网络进行社区划分是一项重要工作。为提高多关系网络上社区发现的准确度,提出一种基于机器学习的多关系网络社区发现算法。定义了在多关系网络上计算不同类型的边在社区划分时所占的权重的方法,设计了在多关系网络上采集训练数据的方法,用采集到的数据训练节点表示模型就可以得到网络中节点的向量表示,在节点的向量表示上使用聚类算法进行社区划分。算法在考虑不同类型边之间的差异的前提下直接在多关系网络上实现社区发现。实验结果显示,该算法在单关系网络和多关系网络上优于近几年的一些算法。

Nodes in the real networks are often connected by multiple types of edges.It is an important task to divide the networks into different communities according to the edges among them.In order to improve the accuracy of community detection in multi⁃relational networks,a machine learning based multi⁃relational network community detection algorithm is proposed.The method of calculating the weight of different types of edges in the community partition is defined.The strategy used to collect the training data on multi⁃relational networks is designed.The collected data is used to train the node representation model,so that the vector representation of every node in the networks can be generated.Then,the clustering algorithm is used to divide the community on nodes′vector representation.The algorithm is used to realize community discovery directly on the multi⁃relational network under the premise of considering the differences between different types of edges.Experimental results show that,in comparison with some algorithms in recent years,the proposed algorithm performs better in both single⁃relational networks and multi⁃relational networks.