Towards Reliable Zero-Shot Crowd Forecasting: Evaluating Time Series Foundation Models for Special Event Pedestrian Forecasting

2026-07-20

AI Summary

This study addresses the challenges of crowd flow forecasting under special events—such as scarce historical data, heterogeneous distributions, and short observation windows—which hinder the reliability of conventional supervised methods and their ability to quantify uncertainty. For the first time, we systematically evaluate the zero-shot probabilistic forecasting capabilities of pretrained time series foundation models for crowd flow prediction, demonstrating that they can deliver real-time, accurate forecasts with calibrated uncertainty estimates without requiring local retraining. We introduce decision-oriented evaluation metrics and practical deployment guidelines, and validate our approach on the SAIL2025 event case study. Our analysis clarifies the operational conditions under which zero-shot forecasting is effective, offering actionable insights for crowd management practitioners.

Abstract

Managing massive crowds during infrequent special events requires reliable real-time pedestrian-flow forecasting to ensure public safety and operational efficiency. However, supervised forecasting methods face limitations in these contexts due to scarce historical data, heterogeneous data distributions, and short in-event observation windows. To effectively support operational decision-making, forecasts should provide not only accurate point estimates but also informative predictive uncertainty. Probabilistic uncertainty quantification plays a critical role in this aspect, particularly capturing sudden volatility and tail risks. This paper investigates pretrained time series foundation models as a lightweight approach for zero-shot probabilistic forecasting without extensive local retraining. Using decision-oriented metrics tailored to short events, we conduct a comprehensive assessment of two time series foundation models on crowd forecasting, with the SAIL2025 event as a use case. We then distill practical insights for crowd managers, specifying when zero-shot forecasts remain operationally reliable.

Problem

• zero-shot forecasting
• crowd forecasting
• predictive uncertainty
• special events
• time series foundation models

Innovation

• zero-shot forecasting
• time series foundation models
• probabilistic uncertainty quantification
• crowd forecasting
• special event management