This project aims to develop a graph-based machine learning framework for detecting anomalous behaviour in financial markets, with a primary focus on the US stock market and optional extension to cryptocurrency markets.

The project will model market activity as a graph, where nodes represent financial assets (e.g., US equities or cryptocurrencies) and edges capture relationships such as abnormal correlations, co-movements, or unusual trading activity. By analysing the evolving structure of these graphs over time, the system aims to identify anomalous patterns that may indicate market irregularities, such as coordinated movements or abnormal volatility events.

Historical price and volume data from major US stock indices (e.g., S&P 500, NASDAQ) will be used as the primary dataset. If feasible, the methodology will be extended to cryptocurrency markets (e.g., Bitcoin and Ethereum) to compare anomaly characteristics across traditional and decentralised financial systems.

The project will apply graph-based metrics and machine learning techniques for anomaly detection and evaluate performance using statistical validation, visual analysis, and comparison across market types. This work combines data science, machine learning, and network analysis to provide insights into abnormal market behaviour across different financial environments.