Pathfinder: A Bayesian Network for Lymph Node Pathology Diagnosis

109-node Bayesian network modeling lymph node pathology, developed as part of the Pathfinder expert system project. The network contains 195 arcs and 72,079 parameters, with an average Markov blanket size of 3.82. Authored by D. Heckerman, E. Horwitz, and B. Nathwani, this dataset is a seminal benchmark for probabilistic reasoning, originally published in 1992.

Use Cases

• Benchmarking inference algorithms for Bayesian networks based on the network's 109 nodes and 195 arcs.
• Evaluating the performance of structure learning algorithms against a known, large-scale reference model.
• Studying the application of probabilistic expert systems in medical diagnosis, specifically for lymph node pathology.
• Teaching concepts of probabilistic graphical models using a historically significant, documented network.

Strengths

• Well-documented, seminal network with 109 nodes and 195 arcs providing a substantial benchmark.
• Network parameters (72,079) and structural metrics (average degree 3.58, max in-degree 5) are explicitly quantified.
• Clear academic provenance with a defined citation to the original 1992 research publication.

Limitations

• Column-level documentation and sample data are unavailable, requiring semantic inference after download.
• Row count and file formats are unknown, which may limit suitability assessment for specific tasks.
• Last update date is unknown; the dataset reflects knowledge from its 1992 publication date.

Provenance

Source

D. Heckerman, E. Horwitz, and B. Nathwani; bnlearn Bayesian Network Repository.

Collection Method

Constructed for the Pathfinder expert system project in medical diagnosis.

Time Range

Model reflects knowledge circa 1992.

Freshness

Publication date is 1992; last update date is unknown.

Geography

Not specified.