Semi-Supervised Community Detection via Constraint Matrix Construction and Active Node Selection

Suqi Zhang,Junhua Gu,Jianxin Li,Junyan Wu,Xinyun Xu,Xianchao Tang

doi:10.1109/access.2019.2962634

Abstract

Identification of community structures is essential for characterizing and analyzing complex networks. Having focusing primarily on network topological structures, most existing methods for community detection ignore two types of non-topological relationships among nodes, i.e., pairwise “must-link” constraints among pairs of nodes and labels of nodes, such as functions they may have. Here, we present a novel semi-supervised and active learning method for community detection to integrate these two types of information of a network so as to increase the accuracy of community identification. Our new method will honor the “must-link” relationship without introducing new parameters and is efficient with a guaranteed convergence. An essential component of the method is a linear representation that is particularly suited to an active learning to help select the most critical nodes that impact community discovery. We present results from extensive experiments on synthetic and real networks to show the superior performance of the new methods over the existing approaches.

Highlights

Networks in real world are not random, but rather contain groups or community structures, which manifest organizational structures and functional components of the underlying systems
In this paper we propose a novel, NMFbased, semi-supervised and active learning method, to exploit non-topological information for community detection
BASELINES FOR COMPARISON We experimentally evaluated our methods, named as supervised symmetric nonnegative matrix factorization method (SSNMF) for the basic algorithm and SSNMF_AL for the method with active learning, by comparing them with four state-of-theart community detection methods

Summary

INTRODUCTION

Networks in real world are not random, but rather contain groups or community structures, which manifest organizational structures and functional components of the underlying systems. This method actively selects a small amount of links as side information to ‘‘sharpen’’ the boundaries; refers the number of communities automatically This method transfers the network topology into similarity space for node representation, VOLUME 8, 2020 ,the feature dimensionality will be too high as an increase of nodes in real networks. All of these methods rely on prior information excessively To this end, in this paper we propose a novel, NMFbased, semi-supervised and active learning method, to exploit non-topological information for community detection. In this paper we propose a novel, NMFbased, semi-supervised and active learning method, to exploit non-topological information for community detection It guarantees to abide by the additional constraints on links and nodes. We close with some conclusions in the last Section V

RELATED WORK

MODEL OVERVIEW

CONSTRAINT MATRIX CONSTRUCTION IN SSNMF

COMPLEXITY

Findings

CONCLUSION