萌妹社区

Professor Chulwoo Han from Sungkyunkwan University recently published innovative research on pairs trading using unsupervised learning. Pairs trading is a market-neutral investment strategy that buys an undervalued stock and sells an overvalued stock when two stocks with similar characteristics diverges from each other. This study aims to significantly improve traditional pairs trading strategies to achieve higher returns and expand their applicability in financial markets.

Professor Han has made significant contributions to the burgeoning field of financial research utilizing machine learning. While most existing studies rely on supervised learning, Professor Han explored pairs trading strategies using unsupervised learning in this study. Unsupervised learning clusters data to group samples with similar characteristics, allowing for more accurate identification of stock pairs by considering not only traditional price data but also firm characteristics.

Professor Han and his research team applied prominent clustering algorithms, such as k-means, DBSCAN, and agglomerative clustering, to the U.S. stock market and tested pairs trading strategies. The results show that long-short portfolios created using stock pairs selected through agglomerative clustering achieved an average annual return of 24.8% and a Sharpe ratio of 2.69. This performance is substantially superior to that of traditional pairs trading strategies. Notably, this study demonstrated that the strategy maintained high profitability even after accounting for transaction costs.

“Identifying stock pairs taking firm characteristics into account provides much higher accuracy and profitability compared to simple price data-based methods. This research will significantly expand the possibilities of unsupervised learning in financial markets." said Professor Han. The research team confirmed through various robustness tests that these results were not due to data biases or chance. Furthermore, the strategy using clustering algorithms demonstrated high profitability even in extreme market conditions, such as financial crises.

Professor Han expects this research to make significant contributions to the fields of financial engineering and machine learning. "We will be able to develop more sophisticated and efficient financial models through unsupervised learning," he said. "We will continue our research to develop various algorithms that can be used in actual financial markets." This research is expected to help pursue both stability and profitability in financial markets.

Professor Chulwoo Han's research at Sungkyunkwan University has created a significant impact on both academic and industrial sectors in financial engineering and machine learning, providing important insights for future research directions.