Insurance Bayesian Network Sample 3: A 27-Node Probabilistic Model

Insurance_3 is a discrete Bayesian network sample from the bnlearn repository, containing 27 nodes connected by 52 arcs. The network, authored by J. Binder, D. Koller, S. Russell, and K. Kanazawa, has 1008 parameters and is referenced in their 1997 Machine Learning paper on adaptive probabilistic networks with hidden variables.

Use Cases

• Benchmarking Bayesian network learning algorithms based on the provided network structure and parameters.
• Studying conditional independence relationships and Markov blankets based on the 27-node, 52-arc topology.
• Simulating insurance-related risk scenarios using the probabilistic model described in the repository.
• Teaching concepts of directed acyclic graphs and probabilistic reasoning in an insurance domain context.

Strengths

• Well-defined network structure with 27 nodes and 52 arcs, providing a concrete probabilistic model.
• Detailed topological metrics are provided, including an average Markov blanket size of 5.19 and a maximum in-degree of 3.
• Clear academic provenance with a citable 1997 paper from known authors in the machine learning field.

Limitations

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

Provenance

Source

bnlearn Bayesian Network Repository, sample contributed by J. Binder, D. Koller, S. Russell, and K. Kanazawa.

Collection Method

Likely a synthetic or expert-constructed Bayesian network for research and benchmarking purposes.

Time Range

null

Freshness

unknown

Geography

null