Portfolio Optimization Adapted to User Preference for Risk Using Preference-Aware Multi-Objective Bayesian Optimization [in Japanese]

Ryota Ozaki, Masanori Hirano

The 36th meeting of Special Interest Group on Financial Informatics of Japanese Society for Artificial Intelligence, pp. 99-105, Mar. 21, 2026

Conference

The 36th meeting of Special Interest Group on Financial Informatics of Japanese Society for Artificial Intelligence (SIG-FIN)

Abstract

In portfolio optimization for equity investment, user preferences are one of the important factors. For example, it is important to reflect each investor's preference for risk-return trade-offs. However, it is generally difficult for investors to directly specify preference parameters such as risk aversion as concrete values. This paper proposes a method that uses preference feedback from investors (pairwise comparisons and improvement direction requests) to sequentially estimate preferences while searching for risk aversion parameters in the mean-variance model, thereby efficiently reaching solutions that achieve desirable trade-offs for users. Numerical experiments using Japanese stock data show that the proposed method can find more desirable solutions with fewer exploration steps.

Keywords

Portfolio Optimization; Multi-Objective Bayesian Optimization; User Preference;

doi

10.11517/jsaisigtwo.2026.FIN-036_99

bibtex

@inproceedings{Ozaki2026-sigfin36,
  title={{Portfolio Optimization Adapted to User Preference for Risk Using Preference-Aware Multi-Objective Bayesian Optimization [in Japanese]}},
  author={Ryota Ozaki and Masanori Hirano},
  booktitle={The 36th meeting of Special Interest Group on Financial Informatics of Japanese Society for Artificial Intelligence},
  pages={99-105},
  doi={10.11517/jsaisigtwo.2026.FIN-036_99},
  year={2026}
}