Mining association rules with improved semantics in medical databases

The discovery of new knowledge by mining medical databases is crucial in or der to make an effective use of stored data, enhancing patient management t asks. One of the main objectives of data mining methods is to provide a cle ar and understandable description of patterns held in data. We introduce a new approach to find association rules among quantitative values in relatio nal databases. The semantics of such rules are improved by introducing impr ecise terms in both the antecedent and the consequent, as these terms are t he most commonly used in human conversation and reasoning. The terms are mo deled by means of fuzzy sets defined in the appropriate domains. However, t he mining task is performed on the precise data. These "fuzzy association r ules" are more informative than rules relating precise values. We also intr oduce a new measure of accuracy, based on Shortliffe and Buchanan's certain ty factors [Shortliffe E, Buchanan B. Math Biosci 1975;23:351-79]. Also, th e semantics of the usual measure of usefulness of an association rule, call ed support are discussed and some new criteria are introduced. Our new meas ures have been shown to be more understandable and appropriate than ordinar y ones. Several experiments on large medical databases show that our new ap proach can provide useful knowledge with better semantics in this field. (C ) 2001 Elsevier Science B.V. All rights reserved.