The UCI Heart Disease Dataset is a heart disease dataset that contains a total of 76 attributes, but all published experiments refer to a subset of 14 of them, with the Cleveland database being the only database that ML researchers have used. The "goal" field refers to whether the patient has heart disease, and the experiments in the Cleveland database mainly focus on trying to distinguish between presence (values 1, 2, 3, 4) and absence (value 0).

The UCI Heart Disease dataset was published in 1989 by:

• Hungarian Institute of Cardiology. Budapest: Andras Janosi, MD
• University Hospital, Zurich, Switzerland: William Steinbrunn, MD
• University Hospital, Basel, Switzerland: Matthias Pfisterer, MD
• VA Medical Center, Long Beach and Cleveland Clinic Foundation: Robert Detrano, MD, Ph.D

Related papers include:

• "International application of a new probability algorithm for the diagnosis of coronary artery disease. American Journal of Cardiology"
• "Instance-based prediction of heart-disease presence with the Cleveland database"
• "Models of incremental concept formation. Artificial Intelligence"