Using machine learning for stock picking

Description

Gaining exposure to small-cap companies may appeal to investors for a number of reasons. These companies have shown strong historic returns compared to their large-cap counterparts and offer a lower risk for longer investment horizons. Moreover, the relative scarcity of small-cap-focused funds compared to the propensity of stocks available signifies greater opportunity in a stock universe that tends to be less thoroughly researched.

Even after considering the various constraints in choosing companies eligible for inclusion in such a portfolio—whether these be business fundamental (e.g., market cap or liquidity), geographical (e.g., excluding emerging markets), or ethical (e.g., excluding companies based on the United Nations Global Compact) considerations—the available investment universe is huge.

Alun Biffin and David Dogon explain how machine learning methods can revolutionize the stock-picking process by processing vast amounts of fundamental and technical data as well as leveraging novel data sources to provide intelligent stock recommendations. By algorithmically empowering this process of filtering thousands of small-cap stocks down to a handful of potential candidates, portfolio managers can devote their time to carry out the high-level business analysis required before deciding to include a stock in their portfolio. Moreover, the volume and variety of data used to train the model means that portfolio managers are alerted to investment opportunities that might otherwise have laid outside their usual field of research.