Hailfinder: A Bayesian Network for Severe Weather Forecasting

56-node Bayesian network designed for forecasting severe weather, specifically hail. The model contains 66 arcs and 2656 parameters, with an average Markov blanket size of 3.54. It was authored by B. Abramson, J. Brown, W. Edwards, A. Murphy, and R. L. Winkler and published in the International Journal of Forecasting in 1996.

Use Cases

• Benchmarking Bayesian network inference algorithms based on the 56-node, 66-arc structure.
• Studying probabilistic reasoning for weather forecasting based on the model's domain.
• Teaching concepts of graphical models and parameter learning based on the network's documented size and complexity.
• Comparing structure learning algorithms against a known ground-truth network.

Strengths

• Well-documented structure with 56 nodes and 66 arcs providing a clear benchmark topology.
• Specific parameter count (2656) and network statistics (average degree 2.36, maximum in-degree 4) are provided for reproducibility.
• Clear academic provenance with a cited 1996 publication in a peer-reviewed journal.

Limitations

• Column-level documentation and sample data are unavailable; field semantics must be inferred after download.
• Row count and file formats are unknown, which may limit suitability assessment for certain tasks.
• Last update date is unknown; freshness unverified.

Provenance

Source

bnlearn Bayesian Network Repository, referencing academic publication by Abramson et al.

Collection Method

null

Time Range

null

Freshness

null

Geography

null