Network Embedding based on Community- aware and Random Walk for Community Detection in Attributed Networks

Community detection aims to find a collection of closely connected nodes in a network, which is an important problem in network analysis. In recent years, network embedding techniques based on random walks have been widely applied to community detection tasks. However, these methods suffer from problems such as ignoring node attributes and locations, as well as the tediousness of manually setting the walking parameters. To address these issues, this paper proposes a network embedding based on community-aware and random walk for community detection in attributed networks (RWACD). Firstly, we calculate the node topology and attribute similarity to construct the fusion similarity matrix. Secondly, we construct transition probability matrices biased towards community boundary and internal walking, respectively, to enhance nodes' walking inside and at the boundary of the community and to better recognize the community structure. Next, we set the number and length of walking according to the node degree and the number of attributes to simplify the process of setting the walking parameters under the premise of ensuring the accuracy of the algorithm and avoiding the problems of overfitting caused by manually setting the walking parameters. Finally, we get the embedding vectors of nodes by Skip-Gram model and apply k-means clustering to get the community detection results. The effectiveness of the proposed algorithm is verified by conducting experiments on real-world and synthetic datasets.