Arbitrage Detection in Crypto Markets Using Graph Neural Networks

DOI

10.2991/978-94-6463-866-0_8How to use a DOI?

Keywords

cryptocurrency arbitrage; graph neural networks; graphSAGE; edge feature fusion; Bellman-Ford; random walk cycle scanner; real-time arbitrage detection

Abstract

Cryptocurrency arbitrage offers profit opportunities through price discrepancies across exchanges, but identifying viable paths is challenging due to the volatility, fees, and complexity of multi-token ecosystems. This work proposes a scalable and interpretable GNN-based framework using GraphSAGE with custom edge fusion to detect arbitrage opportunities across multiple centralized exchanges. We constructed graph snapshots from 200 five-minute intervals using real trading data from KuCoin, Gate.io, Huobi, Bitget, and MEXC, with six major cryptocurrencies (BTC, ETH, SOL, XRP, LTC, ADA). Edge features include-log (exchange rate), inverse rate, volume, volatility, trading fee, and one-hot encoded exchange identifiers to enhance path-level reasoning. Our model achieved an average F1-score of 0.90, precision of 0.89, recall of 0.92, and AUC of 0.94 across all snapshots, outperforming classical Bellman-Ford cycle detection and a Random Walk-based cycle scanner. We also present computational benchmarks, with average inference time per graph at 78ms on CPU, demonstrating feasibility for real-time systems. We discuss the risks of slippage, latency, and transaction fees in live deployment and suggest mitigation strategies. Our results confirm that a structural GNN approach with rich edge fusion provides an effective balance between performance, interpretability, and deployment readiness for arbitrage detection in crypto markets. This work demonstrates a practical GNN-based arbitrage system with real-time feasibility and measurable financial gains.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Yuchitra Venkatesh
AU  - T. Anusha
AU  - Shwetha Manoj
AU  - P. Vidya
PY  - 2025
DA  - 2025/10/31
TI  - Arbitrage Detection in Crypto Markets Using Graph Neural Networks
BT  - Proceedings of the International Conference on Intelligent Systems and Digital Transformation (ICISD 2025)
PB  - Atlantis Press
SP  - 67
EP  - 76
SN  - 2589-4919
UR  - https://doi.org/10.2991/978-94-6463-866-0_8
DO  - 10.2991/978-94-6463-866-0_8
ID  - Venkatesh2025
ER  -