bnlearn datasets (Q4834873): Difference between revisions

The bnlearn dataset collection consists of five structure learning datasets from the Bayesian Network Repository. These datasets are intended for studying causal discovery algorithms, with known or estimated ground truth structures and exclusively discrete variables. The Alarm dataset is a medical diagnosis dataset with 37 variables, including central venous pressure (CVP), pulmonary capillary wedge pressure (PCWP), and total peripheral resistance (TPR). The Asia dataset is a synthetic binary dataset modeling respiratory diseases, containing variables such as tuberculosis, lung cancer, smoking, and chest X-ray results. The Coronary dataset focuses on risk factors for coronary heart disease, including smoking, strenuous physical and mental work, systolic blood pressure, and family history. The Hailfinder dataset includes 56 meteorological variables related to hail forecasting, such as vertical motion, meso-alpha area, and satellite moisture contributions. The Lizards dataset contains three variables describing lizard species, perch height, and perch diameter. Each dataset provides a well-defined structure for evaluating causal discovery methods, making them valuable for Bayesian network research and algorithm development.