Spatiotemporal forecasting benchmarks criticized for bias, favoring linear models

A new paper critically examines common benchmark datasets and baselines used for spatiotemporal forecasting with graph neural networks (GNNs). The research highlights that spatially unaware linear models often perform competitively against GNNs on widely adopted datasets like Chickenpox, METR-LA, and PEMS-BAY. The authors attribute this to structural biases in the datasets and propose a more rigorous statistical evaluation methodology to better assess GNN performance and identify novel model development approaches. AI

IMPACT Highlights potential overestimation of GNN performance in spatiotemporal forecasting, urging for more rigorous evaluation.

RANK_REASON Academic paper analyzing existing research methodologies and datasets. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv stat.ML →

• arXiv
• Chickenpox
• graph neural networks
• METR-LA
• PedalMe
• PEMS-BAY
• WikiMaths

• paper
• other

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