CHAPTER 7Ensemble Learning Applied to Quant Equity: Gradient Boosting in a Multifactor Framework

Tony Guida and Guillaume Coqueret


It is a both intuitive and well‐documented fact that firms' performance on the stock market is driven by some of their core characteristics. In their seminal article, Fama and French (1992) show that firms with higher book‐to‐market ratios significantly outperform those with low book‐to‐market ratios. They also report that small firms tend to yield returns that are higher than those of large firms.1 Later, Jegadeesh and Titman (1993, 2001) constructed abnormally profitable (momentum) portfolios by buying outperforming stocks and shorting underperforming ones.

Findings such as these have led to the construction of so‐called factor indices in which the investor buys the above‐average performing stocks and sells the below‐average ones. The literature on these anomalies is incredibly vast and has its own meta‐studies (see e.g. Subrahmanyam 2010; Green et al. 2013; Harvey et al. 2016).2

It can be debated whether these discrepancies in performance originate from truly pervasive (and priced) factors that structure the cross‐section of stock returns (a stream of literature that was launched by Fama and French 1993) or from the firms' characteristics directly, as put forward by Daniel and Titman (1997).

In any case, there seems to be a large consensus that investors should be able to benefit from the introduction of firms' characteristics ...

Get Big Data and Machine Learning in Quantitative Investment now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.