Time Series Foundation Models are Flow Predictors

TL;DR

This paper demonstrates that time series foundation models like Moirai and TimesFM can effectively predict crowd flow in various real-world datasets without explicit spatial data, outperforming traditional methods.

Contribution

It introduces the application of TSFMs to crowd flow prediction in a zero-shot setting, showing significant improvements over existing statistical and deep learning models.

Findings

Moirai and TimesFM outperform baselines with up to 33% lower RMSE.

Achieve 39% lower MAE and 49% higher CPC than competitors.

Effective even with limited data or missing spatial information.

Abstract

We investigate the effectiveness of time series foundation models (TSFMs) for crowd flow prediction, focusing on Moirai and TimesFM. Evaluated on three real-world mobility datasets-Bike NYC, Taxi Beijing, and Spanish national OD flows-these models are deployed in a strict zero-shot setting, using only the temporal evolution of each OD flow and no explicit spatial information. Moirai and TimesFM outperform both statistical and deep learning baselines, achieving up to 33% lower RMSE, 39% lower MAE and up to 49% higher CPC compared to state-of-the-art competitors. Our results highlight the practical value of TSFMs for accurate, scalable flow prediction, even in scenarios with limited annotated data or missing spatial context.