Improving Micro-Extended Belief Rule-Based System Using Activation Factor for Classification Problems

Long-Hao Yang, Jun Liu, Ying-Ming Wang, Hui Wang, Luis Martínez

Abstract: The micro-extended belief rule-based system (Micro-EBRBS) is an advanced rule-based system and has shown its superior ability in solving big data problems. To overcome the activation rule incompleteness and inconsistency of Micro-EBRBS, a new concept, named activation factor (AF), is introduced to revise the calculation of individual matching degree and, furthermore, an AF-based inference (AFI) method is proposed for improving Micro-EBRBS. A comparative analysis study is conducted using three classification datasets. Results demonstrate that the proposed AFI method can not only improve the accuracy of Micro-EBRBS, but also reduce the number of failed data in the process of rule inference.