Machine Learning and Portfolio Management: A review

Abstract

Linkages across different asset classes and relationships of innovative financial products such as cryptocurrencies and global macroeconomic events have made price and return predictions more challenging.  This paper investigates how Machine Learning (ML) technologies can be implemented in two critical areas of active portfolio management.  Specifically, we focus on ML techniques which can be used in (i) fundamental and technical analysis, to make better asset allocation decisions; and (ii) portfolio construction, to make more efficient and optimized portfolios.  We surveyed the existent literature, including trends of ML in finance, and the main benefits and challenges such technologies can bring to the portfolio management industry.  Existing ML applications’ benefits outweigh some costs in terms of human resources and efficiency in financial decision making, with however some key challenges in data quality, including understanding and interpretation of ML models which need to be addressed.