FinCast: A Foundation Model for Financial Time-Series Forecasting

2025-08-27

AI Summary

Financial time series forecasting faces three core challenges: temporal non-stationarity, cross-domain heterogeneity (e.g., equities, commodities, futures), and multi-temporal-resolution modeling (from seconds to weeks). Existing deep learning approaches suffer from overfitting and heavy reliance on domain-specific fine-tuning. This paper introduces FinCast—the first foundation model for financial time series forecasting. FinCast is pre-trained on large-scale, heterogeneous financial data and integrates multi-scale temporal modeling with domain-invariant feature learning. Crucially, it achieves zero-shot generalization across markets and asset classes—enabling direct deployment to unseen scenarios without any fine-tuning. Extensive experiments on multiple benchmarks demonstrate that FinCast significantly outperforms state-of-the-art methods, validating its strong robustness and superior zero-shot forecasting capability.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: (Large) Language Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Financial time-series forecasting is critical for maintaining economic stability, guiding informed policymaking, and promoting sustainable investment practices. However, it remains challenging due to various underlying pattern shifts. These shifts arise primarily from three sources: temporal non-stationarity (distribution changes over time), multi-domain diversity (distinct patterns across financial domains such as stocks, commodities, and futures), and varying temporal resolutions (patterns differing across per-second, hourly, daily, or weekly indicators). While recent deep learning methods attempt to address these complexities, they frequently suffer from overfitting and typically require extensive domain-specific fine-tuning. To overcome these limitations, we introduce FinCast, the first foundation model specifically designed for financial time-series forecasting, trained on large-scale financial datasets. Remarkably, FinCast exhibits robust zero-shot performance, effectively capturing diverse patterns without domain-specific fine-tuning. Comprehensive empirical and qualitative evaluations demonstrate that FinCast surpasses existing state-of-the-art methods, highlighting its strong generalization capabilities.

Problem

• Addresses financial time-series forecasting challenges
• Overcomes temporal non-stationarity and multi-domain diversity
• Eliminates need for domain-specific fine-tuning

Innovation

• Foundation model for financial time-series forecasting
• Robust zero-shot performance without fine-tuning
• Trained on large-scale multi-domain financial datasets