Extracting Residual Returns via Principal Component Analysis and Gaussian Graphical Models for a Momentum Strategy [in Japanese]

Koshi Watanabe, Ryota Ozaki, Kentaro Imajo, Masanori Hirano

The 36th meeting of Special Interest Group on Financial Informatics of Japanese Society for Artificial Intelligence, pp. 264-271, Mar. 22, 2026

Conference

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

Abstract

Residual returns are known to be effective for trading strategies that hedge market-wide risks. Multivariate analysis methods are used to identify residual returns, but instability such as rank deficiency, which frequently occurs in financial time series, poses a significant challenge. This paper proposes a method for extracting residual factors by combining principal component analysis and a Gaussian graphical model. We confirm that the residual returns obtained by the proposed method exhibit higher stability and orthogonality compared to conventional PCA-based methods. Furthermore, backtests using historical data of S&P 500 and TOPIX 500 constituents confirm improved orthogonality of residual returns and improved Sharpe ratios in reversal strategies.

Keywords

Residual Returns; Principal Component Analysis; Gaussian Graphical Model; Momentum Strategy;

doi

10.11517/jsaisigtwo.2026.FIN-036_264

bibtex

@inproceedings{Watanabe2026-sigfin36,
  title={{Extracting Residual Returns via Principal Component Analysis and Gaussian Graphical Models for a Momentum Strategy [in Japanese]}},
  author={Koshi Watanabe and Ryota Ozaki and Kentaro Imajo and Masanori Hirano},
  booktitle={The 36th meeting of Special Interest Group on Financial Informatics of Japanese Society for Artificial Intelligence},
  pages={264-271},
  doi={10.11517/jsaisigtwo.2026.FIN-036_264},
  year={2026}
}