Data-Driven Dynamic Factor Modeling via Manifold Learning

Graeme Baker, Agostino Capponi, J. Antonio Sidaoui · 2025

Read Data-Driven Dynamic Factor Modeling via Manifold Learning by Graeme Baker, Agostino Capponi, J. Antonio Sidaoui (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-19945.txt.

Introduction

Graeme Baker, Agostino Capponi, and Antonio Sidaoui build a factor model for markets without fixed formulas, in this 2025 paper. Standard factor models learn from asset prices alone and can lose the link to the outcomes they should predict. Their method learns a low-dimensional map that keeps both the shape of the data and its tie to responses.

For time series driven by diffusion, they prove the learned map converges to the true process and bound its error. They then forecast within the map using Kalman filtering. Applied to bank stress tests with Federal Reserve scenarios, the method cut forecast error well below standard scenario work and a classic benchmark.

Catalog record

CREATORGraeme Baker, Agostino Capponi, J. Antonio Sidaoui
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE145 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-19945

Browse the full library · About the Tower of Babel