Network classification through random walks

2025-05-27

AI Summary

This study addresses the fundamental problem of classifying system types based solely on network topology, without requiring prior knowledge of underlying dynamical models. We propose a topology-driven classification framework that constructs discriminative “network fingerprints” from multi-order random walk statistics—including first-passage time distributions, visit-frequency entropy, and path-length variance—thereby overcoming the representational limitations of conventional single-scale structural metrics. The method integrates random walk sampling, high-order statistical feature extraction, and supervised learning (SVM/Random Forest). Evaluated on seven real-world and synthetic networks (e.g., LFR, Cora, Email-Eu), it achieves an average accuracy improvement of 5.2% over state-of-the-art baselines such as Graph2Vec and NetLSD. Notably, it demonstrates superior robustness on medium-sized sparse networks, validating that random-walk statistics effectively capture system-level topological essence.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization

Abstract

Network models have been widely used to study diverse systems and analyze their dynamic behaviors. Given the structural variability of networks, an intriguing question arises: Can we infer the type of system represented by a network based on its structure? This classification problem involves extracting relevant features from the network. Existing literature has proposed various methods that combine structural measurements and dynamical processes for feature extraction. In this study, we introduce a novel approach to characterize networks using statistics from random walks, which can be particularly informative about network properties. We present the employed statistical metrics and compare their performance on multiple datasets with other state-of-the-art feature extraction methods. Our results demonstrate that the proposed method is effective in many cases, often outperforming existing approaches, although some limitations are observed across certain datasets.

Problem

• Classify network types based on structural features
• Extract features using random walk statistics
• Compare performance with existing feature extraction methods

Innovation

• Uses random walk statistics for network classification
• Compares metrics with state-of-the-art methods
• Outperforms existing approaches in many cases