[Paper Review] A Survey of Community Detection Approaches: From Statistical Modeling to Deep Learning

Di Jin, Zhizhi Yu+6 more|arXiv (Cornell University)|Jan 3, 2021

Complex Network Analysis Techniques160 references53 citations

TL;DR

This survey presents a unified view of community detection methods, dividing them into probabilistic graphical models and deep learning, and provides benchmark datasets and future directions. It highlights theoretical analyses, taxonomy, and practical resources for learning-based approaches.

ABSTRACT

Community detection, a fundamental task for network analysis, aims to partition a network into multiple sub-structures to help reveal their latent functions. Community detection has been extensively studied in and broadly applied to many real-world network problems. Classical approaches to community detection typically utilize probabilistic graphical models and adopt a variety of prior knowledge to infer community structures. As the problems that network methods try to solve and the network data to be analyzed become increasingly more sophisticated, new approaches have also been proposed and developed, particularly those that utilize deep learning and convert networked data into low dimensional representation. Despite all the recent advancement, there is still a lack of insightful understanding of the theoretical and methodological underpinning of community detection, which will be critically important for future development of the area of network analysis. In this paper, we develop and present a unified architecture of network community-finding methods to characterize the state-of-the-art of the field of community detection. Specifically, we provide a comprehensive review of the existing community detection methods and introduce a new taxonomy that divides the existing methods into two categories, namely probabilistic graphical model and deep learning. We then discuss in detail the main idea behind each method in the two categories. Furthermore, to promote future development of community detection, we release several benchmark datasets from several problem domains and highlight their applications to various network analysis tasks. We conclude with discussions of the challenges of the field and suggestions of possible directions for future research.

Motivation & Objective

• Provide a unified overview of learning-based community detection methods.
• Introduce a two-category taxonomy: probabilistic graphical models and deep learning-based approaches.
• Analyze theoretical connections, challenges, and differences between methods.
• Release benchmark datasets to promote future research in community detection.
• Discuss real-world applications and future research directions.

Proposed method

• Classify existing methods into two main categories: probabilistic graphical models and deep learning.
• Detail subcategories within probabilistic models (directed, undirected, and hybrid) and within deep learning (auto-encoder, GAN, GCN, and hybrid with graphical models).
• Describe representative models and learning paradigms (e.g., stochastic block model, MMSB, topic models, matrix factorization, and graph neural networks).
• Provide a unified architectural view of learning-based community detection methods.
• Release benchmark datasets and discuss applications to motivate further research.

Experimental results

Research questions

• RQ1 What are the main learning-based paradigms for community detection and how do they differ in modeling assumptions?
• RQ2 How can probabilistic graphical models and deep learning approaches be organized into a unified taxonomy?
• RQ3 What theoretical insights and practical challenges characterize learning-based community detection?
• RQ4 What benchmark resources and applications help advance future research in the field.

Key findings

• The paper offers the first comprehensive overview of learning-based community detection organized into probabilistic graphical models and deep learning.
• It analyzes similarities, differences, and challenges across methods and proposes five directions for future work.
• A unified system architecture is presented to relate statistical modeling and deep learning approaches.
• Benchmark datasets and applications across domains are released to support future research.
• The survey discusses dynamic, overlapping, and link-community extensions of SBMs and related methods.
• It highlights how deep learning (GCN, auto-encoders, GANs) complements traditional probabilistic modeling in handling high-dimensional network data.