In helping physicians, decision support systems are becoming a very important part of medical decision making. A prospective candidate for such tasks are decision trees that have already been successfully used for many medical decision making purposes. Although very effective and reliable, the traditional construction approach, which hasn't changed much since its introduction, is still not perfect in real-world decision making. Therefore we decided to try something else-evolutionary construction of decision trees based on genetic algorithms. The whole evolution process has been defined, the appropriate internal representation selected, adequate genetic operators developed and a straightforward evaluation function constructed. Several solutions were evolved for the classification of mitral valve prolapse syndrome. A comparison has been made to the traditional induction of decision trees. Regarding all the advantages that our new approach brings and the quality of obtained results, it can be considered as a good choice for all kinds of real-world decision making.
展开▼
