A GANs-Based Approach for Stock Price Anomaly Detection and Investment Risk Management

Abstract

This paper addresses the challenges of risk management in the financial market through a data-driven approach. In investment management, it is important to detect and avoid market anomalies, defined as significant deviations from typical stock price movement patterns. We propose a method that utilizes Generative Adversarial Networks (GANs) to detect anomalous patterns in stock prices. We devise simple investment strategies based on the GANs-based anomaly detection model and validate them on real-world data to demonstrate if the proposed approach can reduce investment risk. Experimental results show that GANs-based anomaly detection can be successfully incorporated into investment strategies to yield superior returns and Sharpe ratio and much reduced volatility and maximum drawdown compared to the S&P500 benchmark. This study shows the potential of data-driven approaches in detecting and managing investment risk.