49,326 cases of building energy data organized for graph-based learning. The dataset includes 5,481 unique buildings and 64 unique weather IDs, with temporal sequences ranging from 968 to 8,760 steps. It was created by ArchEGraph and last updated on 2026-05-07.
Use Cases
- Predict building energy consumption based on weather-conditioned data
- Train graph neural networks on building and geometry relationships
- Analyze energy usage patterns across different building types and sizes
- Benchmark machine learning models for energy forecasting
Strengths
- 49,326 total cases provide a substantial sample size
- 5,481 unique buildings offer diversity in building types
- 64 unique weather IDs allow for varied climate conditioning
- Includes separate geometry files for spatial context
Limitations
- Column-level documentation is absent; field semantics must be inferred after download
- Row count per file is unknown, which may limit suitability assessment
- Description metadata is limited; actual data quality requires manual inspection after download
Provenance
- Source
- ArchEGraph
- Freshness
- Last updated 2026-05-07 08:11:48