Enhancing portfolio resilience to systemic risk: A neural network approach
This paper aims to enhance the classical mean–variance portfolio selection model by using machine learning techniques and accounting for systemic risk. The optimal portfolio is obtained through a three-step supervised learning model. Firstly, the Smooth Pinball Neural Network is employed to predict return distributions of individual assets and the market. Secondly, we use a copula to model dependence between assets and the market, based on which we simulate return scenarios. Lastly, we maximize an ex-ante Sharpe ratio conditioning on systemic events. We train our models using a comprehensive dataset of nearly 600 U.S. individual stocks over 37 years; however, the portfolio analysis and backtesting are conducted on several representative portfolio sets with sizes up to 50 assets. This design allows the models to learn from a broad cross-section of firms while focusing the evaluation on computationally tractable portfolio dimensions. Our set of predictors includes 94 firm characteristics, 14 macroeconomic variables, and 74 industry dummies. The backtesting results demonstrate the superiority of our proposed approach over popular benchmark strategies including a GARCH-based model. This outperformance is statistically significant and robust to the inclusion of transaction costs.
Authors
- Abderrahim Taamouti (ORCID: https://orcid.org/0000-0002-1360-8803)
- Weidong Lin (ORCID: https://orcid.org/0000-0003-0153-5691)
Institutions
- NEOMA Business School (FR)
- University of Liverpool (GB)
- Durham University (GB)
Publication Details
- Journal
- Journal of Empirical Finance
- Published
- 2026-11-18
- DOI
- https://doi.org/10.1016/j.jempfin.2026.101798
- Primary Topic
- Risk and Portfolio Optimization
- Type
- article
- Field-Weighted Citation Impact
- 0.00