Machine learning can be used in many potential places to get a better understanding of financial markets. The most important function regarding the portfolio construction is the return prediction. The research paper ‘Can Machines ‘Learn’ Finance?’ faces a unique set of challenges that differ markedly from other domains where machine learning has excelled. Understanding these differences is critical for developing impactful approaches and realistic expectations for machine learning in asset management.
What Is Machine Learning?
“The definition of ‘machine learning’ is inchoate and is often context specific. We use the term to describe (i) a diverse collection of high-dimensional models for statistical prediction, combined with (ii) so-called ‘regularization’ methods for model selection and mitigation of overfit, and (iii) efficient algorithms for searching among a vast number of potential model specifications.” This definition from Gu, Kelly, and Xiu (GKX, forthcoming) provides a detailed summary of machine learning in their study of financial markets. The approach of machine learning is to maximize predictive accuracy.
Finance Is Different
The research examines a variety of beneficial use cases and potential pitfalls and emphasizes the importance of economic theory and human expertise for achieving success through financial machine learning. Since machine learning has done so many extraordinary advances like recognizing images or speech, “it may seem a foregone conclusion that it will dominate at financial tasks like stock picking.”
- The first challenge in asset management -return prediction- is the small data problem. And the only way to expand the size of data in return modeling is to wait for time to pass.
- The second critical difference between machine learning for asset management and other applications is the low signal-to-noise ratio in returns. The reason for this is that financial market behavior is hard to predict.
- The other challenges contain the evolving character of markets, the unstructured data sources like news articles or image data and the tradeoff between predictable and interpretable machine learning models.
Beyond Return Prediction
Return prediction poses a difficult challenge for machine learning, because of its small data and low signal-to-noise ratios. But transaction cost management and risk management “are less subject to these limitations and in turn can benefit more from machine learning.”
Source: Israel, Ronen and Kelly, Bryan T. and Moskowitz, Tobias J., Can Machines ‘Learn’ Finance? (January 10, 2020).