Re(Visiting) Time Series Foundation Models in Finance
This addresses the challenge of noisy and non-stationary financial data for traders and risk managers, but it is incremental as it focuses on domain-specific adaptation of existing methods.
The paper tackled financial time series forecasting by evaluating time series foundation models (TSFMs) in global markets, finding that models pre-trained from scratch on financial data achieved substantial forecasting and economic improvements, while off-the-shelf pre-trained TSFMs performed poorly.
Financial time series forecasting is central to trading, portfolio optimization, and risk management, yet it remains challenging due to noisy, non-stationary, and heterogeneous data. Recent advances in time series foundation models (TSFMs), inspired by large language models, offer a new paradigm for learning generalizable temporal representations from large and diverse datasets. This paper presents the first comprehensive empirical study of TSFMs in global financial markets. Using a large-scale dataset of daily excess returns across diverse markets, we evaluate zero-shot inference, fine-tuning, and pre-training from scratch against strong benchmark models. We find that off-the-shelf pre-trained TSFMs perform poorly in zero-shot and fine-tuning settings, whereas models pre-trained from scratch on financial data achieve substantial forecasting and economic improvements, underscoring the value of domain-specific adaptation. Increasing the dataset size, incorporating synthetic data augmentation, and applying hyperparameter tuning further enhance performance.