Using quick decision tree algorithm to find better RBF networks

Abstract

It is known that generated knowledge models for data mining tasks are dependent upon supplied data sets, so supplying good data sets for target data mining algorithms is important for the success of data mining. Therefore, in order to find better RBF networks of k-means clustering efficiently, we refer to the number of errors that are from decision trees, and use the information to improve training data sets for RBF networks and we also refer to terminal nodes to initialize the k value. Experiments with real world data sets showed good results. © 2011 Springer Berlin Heidelberg.