Foundation models show mixed results for pedestrian crowd forecasting

A new study published on arXiv evaluates the effectiveness of time-series foundation models (FMs) for pedestrian crowd count forecasting. The research compares seven different forecasting approaches, including traditional methods like Seasonal Naive and gradient-boosted trees, alongside deep learning models and two FMs (TimesFM and Chronos-2). Findings indicate that while FMs perform well in data-rich, seasonal environments with long contexts, simpler models like Seasonal Naive can remain competitive for long-horizon forecasting with limited historical data. The study emphasizes that model selection should consider data characteristics and forecasting horizons. AI

IMPACT Highlights the need to select AI models for pedestrian forecasting based on data availability and desired forecast horizon.

RANK_REASON The cluster contains a research paper published on arXiv detailing a comparative study of AI models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• Catboost
• Chronos-2 Forecasting Model
• ITS
• LightGBM
• Melbourne
• N-HITS
• PatchTST
• SAIL2025
• TimesFM

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →