Time-Series Foundation Model for Value-at-Risk

2024-10-15

arXiv.org

AI Summary

Accurately forecasting Value-at-Risk (VaR) under small-sample regimes remains challenging for traditional financial time series models. Method: This paper pioneers the application of the time-series foundation model TimesFM to multi-quantile VaR estimation (0.01–0.1), fine-tuning it via supervised learning on 19 years of daily returns from the S&P 100 index and its constituents. We rigorously evaluate performance via backtesting—using actual-to-expected ratio and quantile loss—and benchmark against GARCH, GAS, and empirical quantile methods. Contribution/Results: Fine-tuned TimesFM achieves state-of-the-art or optimal VaR forecasting across all quantiles, consistently outperforming traditional models in calibration (actual-to-expected ratio) and matching the best GAS model in quantile loss. In contrast, zero-shot transfer yields markedly inferior results, underscoring fine-tuning as essential to unlock TimesFM’s risk modeling capability. This work establishes a novel paradigm for financial risk measurement grounded in time-series foundation models.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models
• Computer Vision: Large Vision Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

This study is the first to explore the performance of a time-series foundation model for Value-at-Risk (VaR) forecasting. Foundation models, pre-trained on vast and varied datasets, can be used in a zero-shot setting with relatively minimal data or further improved through finetuning. We compare the performance of Google's model, called TimesFM, against conventional parametric and non-parametric models, including GARCH, Generalized Autoregressive Score (GAS), and empirical quantile estimates, using daily returns from the S&P 100 index and its constituents over 19 years. Our backtesting results indicate that in terms of the actual-over-expected ratio, the fine-tuned TimesFM model consistently outperforms traditional methods. Regarding the quantile score loss function, it achieves performance comparable to the best econometric approach, the GAS model. Overall, the foundation model is either the best or among the top performers in forecasting VaR across the 0.01, 0.025, 0.05, and 0.1 VaR levels. Fine-tuning significantly improves accuracy, indicating that zero-shot use is not optimal for VaR forecasting.

Problem

• Time Series Forecasting
• Financial Risk Prediction
• Model Comparison

Innovation

• TimesFM
• Quantile Score Loss Function
• Risk Value (VaR) Prediction