A Network Embedding and Clustering Algorithm for Expert Recommendation Service

Abstract

Network embedding algorithm is dedicated to learning the low-dimensional representation of network nodes. The feature representations can be used as features of various tasks based on graphs, including classification, clustering, link prediction and visualization. Currently, network embedding algorithms have evolved from considering structures only to considering structures and contents both. However, how to effectively integrate the high-order proximity and node content of the network structure is still a problem to be solved. We propose a new network embedding and clustering algorithm in this paper. We obtain the high-order proximity representation of the information network structure, and the fusion node content completes the low-dimensional representation of the node features, so as to complete the network node clustering for the input of the spectral clustering. In order to further verify the value of the algorithm, we apply the clustering results to the field of expert recommendation, and make influence and activity assessments for domain experts to achieve more valuable expert recommendations. The experimental results show that the proposed algorithm will obtain higher clustering accuracy and excellent expert recommendation results.