A PSO-based Generic Classifier Design and Weka Implementation Study

摘要

Aiming at dataset with different feature attributes and multi nominal decision categories, a new design algorithm using PSO for instance-based learning classifier is put forward which can facilitate to make a better classification decision. On the basis of classes and interfaces in Weka platform, this improved algorithm is implemented and added in the platform. Taking two datasets as for test cases, not only is the classifier compared with optimizing different objective functions, but also is evaluated with other classifiers. Comparisons demonstrate that the proposed algorithm can get more effective classifying results and maybe considered a generic classifier.