This event is free and open to all (RSVP). You will receive the webinar link from no-reply@zoom.us upon registration.
Re(Visiting) Time Series Foundation Models in Finance
This talk presents a comprehensive study on using Time Series Foundation Models (TSFMs), Transformer-based models inspired by Large Language Models (LLMs), for forecasting global financial returns. Using a large-scale dataset of daily excess returns across 94 countries over several decades, the presentation introduces a unified evaluation framework to test how TSFMs can be applied in finance under practical settings such as zero-shot forecasting, model fine-tuning, and training on financial data from scratch, while comparing them against widely used statistical and machine-learning baselines. The goal is to provide a clear, structured view of how foundation model approaches can be integrated into financial prediction workflows, what experimental design choices matter, and how these models should be adapted to the unique challenges of financial markets.
Bio: Eghbal Rahimikia is an assistant professor in financial technology (FinTech) at Alliance Manchester Business School, University of Manchester. His research is situated at the intersection of finance, machine learning, and natural language processing, with an emphasis on applying advanced AI techniques to quantitative finance. He is currently working on the development of specialized AI models, including large language models and, more recently, time series foundation models for financial forecasting. His work contributes to the growing field of AI-driven financial analytics and aims to bridge methodological advances in machine learning with practical applications.