Dimensional Balance Improves Large Scale Spatiotemporal Prediction Performance

In three linesarXiv paper proposing an adaptive framework to improve spatiotemporal forecasting by harmonizing spatial and temporal feature representations. Uses low-rank matrix embedding for spatial compression and extended temporal horizon. Demonstrates substantial accuracy gains on urban traffic, meteorology, and epidemic datasets. Code available on GitHub.

Summary generated by Claude — human-verified