Foundation Time-Series AI Model for Realized Volatility Forecasting

2025-05-16

AI Summary

This paper addresses realized volatility forecasting—a core task in financial risk management—by systematically evaluating and enhancing the applicability of the time-series foundation model TimesFM under zero-shot and dynamic settings. To overcome the limitation of pretrained models in capturing non-stationarity and leptokurtic, heavy-tailed volatility dynamics, we propose an incremental fine-tuning paradigm specifically designed for volatility modeling. Our methodology integrates the TimesFM architecture with incremental learning, zero-shot transfer, and rigorous statistical significance testing (Diebold–Mariano and Giacomini–White tests). Empirical evaluation across multiple financial markets demonstrates that the fine-tuned model consistently outperforms classical econometric benchmarks—including GARCH and HAR—in predictive accuracy, while maintaining deployment readiness. The primary contributions are: (i) the first successful adaptation of a time-series foundation model to volatility forecasting; and (ii) the establishment of a scalable, incremental fine-tuning framework tailored to financial time series.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Cognitive Modeling & Cognitive Systems: Other Foundations of Cognitive Modeling & Systems
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Time series foundation models (FMs) have emerged as a popular paradigm for zero-shot multi-domain forecasting. These models are trained on numerous diverse datasets and claim to be effective forecasters across multiple different time series domains, including financial data. In this study, we evaluate the effectiveness of FMs, specifically the TimesFM model, for volatility forecasting, a core task in financial risk management. We first evaluate TimesFM in its pretrained (zero-shot) form, followed by our custom fine-tuning procedure based on incremental learning, and compare the resulting models against standard econometric benchmarks. While the pretrained model provides a reasonable baseline, our findings show that incremental fine-tuning, which allows the model to adapt to new financial return data over time, is essential for learning volatility patterns effectively. Fine-tuned variants not only improve forecast accuracy but also statistically outperform traditional models, as demonstrated through Diebold-Mariano and Giacomini-White tests. These results highlight the potential of foundation models as scalable and adaptive tools for financial forecasting-capable of delivering strong performance in dynamic market environments when paired with targeted fine-tuning strategies.

Problem

• Evaluating foundation models for financial volatility forecasting
• Assessing zero-shot vs fine-tuned TimesFM performance
• Comparing AI models against traditional econometric benchmarks

Innovation

• Time series foundation model for volatility forecasting
• Incremental fine-tuning adapts to financial data
• Outperforms traditional econometric benchmarks statistically