How effective are Machine Learning methods used in active portfolio management?
An empirical study considers the potential benefits and shortcomings of Machine Learning methods used for investment.
Machine Learning (ML) is a branch of artificial intelligence (AI) concerned with the construction of computer algorithms capable of analysing information from a wide range of datasets, at high speed and with minimal supervision. ML tools are powerful, and have the ability to automatically improve as they gain experience. ML’s potential for financial prediction has seen it embraced by asset managers, and its popularity is expected to increase as the technology becomes ever more sophisticated and powerful.
Unsurprisingly, Machine Learning (ML) methods have also attracted considerable interest from academics. For the research paper Machine Learning for Active Portfolio Management, the authors have evaluated overall performance, and the challenges involved with its use.
The study examines how ML can be employed at three specific stages of the active portfolio management process, namely signal generation, portfolio optimisation, and order/trade execution. It observes the advantages ML has over alternative methods, in terms of its speed and efficiency in handling large amounts of data. The study also identifies risk factors such as poor quality data, and the sheer complexity of the information that is produced, and looks ahead to a time when both the use and study of ML may have been re-shaped by regulation.
The study concludes that ML techniques show great promise for active portfolio management but investors should be aware that potential pitfalls exist.
Machine Learning for Active Portfolio Management has been published in The Journal of Financial Data Science.
