Machine Learning Portfolios with Equal Risk Contributions
38 Pages Posted: 8 Aug 2019 Last revised: 10 Apr 2021
Date Written: April 10, 2021
We investigate the use of machine learning (ML) models to forecast stock returns in the Brazilian equity market, using a rich dataset of technical and fundamental indicators. While all ML models we test are able to produce portfolios that outperform the local market, the performance of traditional long-short strategies based on ML return forecasts is hampered by the high volatility of the short portfolios. We show empirically that a simple Equal Risk Contribution (ERC) approach, applied to the long and short components of individual ML long-short strategies, significantly improves their risk-adjusted returns. We pursue this idea further and develop a multi-strategy ERC approach that combines multiple long-short strategies obtained with various ML models, such that (i) the risk contributions of all individual long-short strategies are equal, and (ii) the aggregate risk contribution of all long positions equals that of all short positions. The multi-strategy ERC approach outperforms, on a risk-adjusted basis, all individual ML long-short strategies, as well as strategies based on forecasts from an ensemble of ML models (with or without the use of ERC), and a multi-strategy approach that invests equally in all ML strategies.
Keywords: machine learning, stock market prediction, portfolio optimization, equal risk contribution
JEL Classification: C53, G11, G15
Suggested Citation: Suggested Citation