Multi-Hypothesis Prediction for Portfolio Optimization: A Structured Ensemble Learning Approach to Risk Diversification

Alejandro Rodriguez Dominguez, Muhammad Shahzad and Xia Hong

Papers from arXiv.org

Abstract: A framework for portfolio allocation based on multiple hypotheses prediction using structured ensemble models is presented. Portfolio optimization is formulated as an ensemble learning problem, where each predictor focuses on a specific asset or hypothesis. The portfolio weights are determined by optimizing the ensemble's parameters, using an equal-weighted portfolio as the target, serving as a canonical basis for the hypotheses. Diversity in learning among predictors is parametrically controlled, and their predictions form a structured input for the ensemble optimization model. The proposed methodology establishes a link between this source of learning diversity and portfolio risk diversification, enabling parametric control of portfolio diversification prior to the decision-making process. Moreover, the methodology demonstrates that the diversity in asset or hypothesis selection, based on predictions of future returns, before and independently of the ensemble learning stage, also contributes to the out-of-sample portfolio diversification. The sets of assets with more diverse but lower average return predictions are preferred over less diverse selections. The methodology enables parametric control of diversity in both the asset selection and learning stages, providing users with significant control over out-of-sample portfolio diversification prior to decision-making. Experiments validate the hypotheses across one-step and multi-step decisions for all parameter configurations and the structured model variants using equity portfolios.

Date: 2025-01, Revised 2025-01
New Economics Papers: this item is included in nep-rmg
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
http://arxiv.org/pdf/2501.03919 Latest version (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:arx:papers:2501.03919

Access Statistics for this paper

More papers in Papers from arXiv.org
Bibliographic data for series maintained by arXiv administrators ().