基于node2vec的社区检测方法

A Community Detection Method Based on Node2vec

社区检测可以帮助分析及预测整个网络各元素间的交互关系,为了进一步提高社区检测的准确度,论文提出了一种基于node2vec的社区检测方法。该方法首先采用一种二阶的随机游走策略生成一系列线性序列,然后使用Skip-Gram模型去训练特征向量,最后使用聚类算法对训练出的节点特征向量进行聚类,实现社区的划分。该文在具有社区标签的网络中进行了实验,从实验中验证了这种思想的可行性,从而取得了显著的效果。

Community detection can help us analyze and predict the interaction between elements of the entire network,in order to improve the accuracy of community detection,a community detection method based on node2vec is proposed. This method firstly uses a second-order random walk strategy to generate a series of linear sequences,then uses the Skip-Gram model to train feature vectors,and finally uses a clustering algorithm to cluster the nodal feature vectors and realizes the division of the community.Experiments are carried out in a network with community tags,and the feasibility of this idea from experiments is verified,and remarkable results are achieved.