A deep learning approach for semi-supervised community detection in Online Social Networks

Social Network Analysis (SNA) has gained popularity as a way to unveil and identify useful social patterns as communities among users. However the continuous, exponential growth of these networks (both in terms of number of users, and in terms of the variety of different interactions that these networks allow) has made the development of efficient and effective community detection techniques a challenging computational task. In this paper, we propose an innovative approach for Semi-supervised Community Detection, exploiting Convolutional Neural Networks to simultaneously leverage different properties of a network - such as topological and context information. Crucially, computational cost is optimized by building on the insight that representing network connections over particular sparse matrices can significantly decrease the number of operations that need to be explicitly performed. By extensively evaluating our system on large (artificial and real-world) datasets, we show that our approach outperforms a variety of existing state-of-the-art techniques in terms of running time, as well as over Macro- and Micro - F1. (C) 2021 Elsevier B.V. All rights reserved.

Automated summary

This paper introduces an innovative approach for semi-supervised community detection using Convolutional Neural Networks, optimizing computational cost and outperforming existing techniques on large datasets in terms of running time and F1 scores.