analiza aiqbench
This research addresses the fundamental tension between model complexity, training window length, and market non-stationarity in financial time series prediction. By proposing an adaptive selection procedure with formal guarantees, it demonstrates that jointly optimizing these parameters yields significant out-of-sample performance gains over fixed-window benchmarks during economic downturns. The work advances machine learning theory by providing a rigorous framework for handling regime shifts without relying on outdated data or overly rigid models.
The Nonstationarity-Complexity Tradeoff in Return Prediction
arXiv:2512.23596v2 Announce Type: replace-cross Abstract: Does more data improve return prediction? In non-stationary financial markets, longer training windows improve prediction of complex models but incorporate outdated economic regimes, whereas simpler models require less data and are less vulnerable to changes in economic conditions. We formally characterize this nonstationarity-complexity tradeoff, showing that model complexity and training window length must be jointly optimized. We propose an adaptive selection procedure with formal performance guarantees. Over three decades of U.S. equity markets, our method improves out-of-sample $R^2$ on industry portfolios by 14% relative to fixed-window and regime-switching benchmarks, with large gains during recessions.
Abstract:Does more data improve return prediction? In non-stationary financial markets, longer training windows improve prediction of complex models but incorporate outdated economic regimes, whereas simpler models require less data and are less vulnerable to changes in economic conditions. We formally characterize this nonstationarity-complexity tradeoff, showing that model complexity and training window length must be jointly optimized. We propose an adaptive selection procedure with formal performance guarantees. Over three decades of U.S. equity markets, our method improves out-of-sample $R^2$ on industry portfolios by 14% relative to fixed-window and regime-switching benchmarks, with large gains during recessions.
Machine Learning (stat.ML); Machine Learning (cs.LG); General Finance (q-fin.GN)
arXiv:2512.23596 [stat.ML]
(or arXiv:2512.23596v2 [stat.ML] for this version)
https://doi.org/10.48550/arXiv.2512.23596 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 192 cuvinte
Subiecte conexe
Policy Complexity, Reaction Time, and Bounded Rationality in Reinforcement Learning
arXiv · 26 sept. 2026
Transactional Vs Analytical Database: Choosing OLTP, OLAP, or Hybrid
Databricks · 24 aug. 2026
How can LLM RL Work Despite Information-Theoretic Inefficiency
LessWrong · 16 oct. 2026
Humans Are Not Chinchillas: Revisiting “How Quick and Big Would A Software Intelligence Explosion Be?”
LessWrong · 5 oct. 2026