In this paper,we apply support vector machine (SVM) to knowledge discovery (KD) and confirm its effectiveness with a benchmark data set.SVM has been successfully applied to problems in various domains.however,its effectiveness as a KD method is unknown.We propose SVM for KD,which deals with a classfication problem with a can sort attributes with respect to their efrfectiveness in discriminating classes.moreover,SVM for KD can discover crucial examples for discrimination.We settled six discovery tasks with the meningoencephalitis the discovery outcomes of SVM for KD from one to five with respect to several criteria.Delectecd attributes in six tasks are all valid and useful: their average scores are 3.8-4.0.Discovering order of attributes about usefulness represents a challenging problem.However,concerning this problem,our method achieved a scored of more than or equal to 4.0 in three tasks.Besides,crucial examples for discrimination and typical examples for each class agree with medical knowledge.These promising results demonstrate the effectiveness for our approach.
展开▼
