Advanced dynamic selection of diagnostic methods

Several data mining methods have recently been developed to extract knowledge from large databases. The problem of selecting the most appropriate data mining method(s) has long been solved using static selection methods, and it is only recently that several effective dynamic selection approaches have been proposed. It is expected that dynamic selection which takes into account the expertise areas of each method will lead to better data mining results. This paper analyzes a method for the dynamic selection of diagnostic methods. This method is proposed for use in an intelligent medical diagnostic system. Real-world medical data is often heterogeneous, containing many cases and attributes, and it needs different processing methods for different cases. The dynamic selection method was tested using three databases included in the University of California Machine Learning Repository, achieving promising results in diagnostic accuracy and/or in the time requirements of diagnostics