Multi-class and multi-attribute are two important features of classification problems and have different effects on the requirements and performance of the classifier. Decomposition strategy and overlap function are two effective ways to enhance the performance of classifiers, because the former decomposes a complex multi-class problem into multiple simple sub-problems; the latter uses various functions to specify the conjunctive relationship of input variables in a multi-attribute problem. Extended belief rule-based system (EBRBS) is an advanced rule-based system that has been widely used in classification problems. In order to apply decomposition strategies and overlap functions to enhance the performance of EBRBSs, the present work focuses on the investigative research and comparative evaluation of the commonly used one-versus-one (OVO) decomposition strategy and five common overlap functions to improve the performance of EBRBSs on multi-class and multi-attribute problems. More specifically, three typical kinds of EBRBSs, namely original EBRBS (O-EBRBS), EBRBS with dynamic rule activation (DRA-EBRBS), and a latest EBRBS for big data (Micro-EBRBS), are selected to conduct extensive experimental studies on twenty classification problems. To best of our knowledge, this present work is the first time to provide a meaningful and useful study in revealing the potential capability of the EBRBSs with decomposition strategy and overlap function for multi-class and multi-attribute problems. Experimental results demonstrate that the square product overlap function and the OVO strategy can enhance the performance of EBRBSs over others for twenty classification problems.
|Journal||International Journal of Machine Learning and Cybernetics|
|Publication status||Published - 14 Jun 2021|