A novel rule generation and activation method for extended belief rule-based system based on improved decision tree

Among many classifiers applied to classification problems, the extended belief rule-based (EBRB) system is a powerful tool with the ability to handle both qualitative and quantitative information under uncertainty. However, it may face the problems of low inference efficiency and inconsistent rule activation in some applications due to the limitations of the conventional EBRB generation and activation method. To this end, a novel rule generation and activation method for EBRB system based on improved decision tree is proposed in this paper. Firstly, the distributed EBRB is generated by using the improved decision tree construction method designed in this paper. Then, the conventional EBRB activation scheme is modified by selecting input-related sub-EBRB for a given input as inference rule base to search suitable belief rules as well as introducing the attribute sorting knowledge in the process of activation. Moreover, the original procedures of the inference process and class estimation are retained from conventional EBRB system. Experiments were carried out to validate the efficiency and effectiveness of the proposed method in comparison with different types of classifiers under eleven standard classification datasets. The comparison results show that the proposed method could obtain satisfactory classification accuracy and rule inference efficiency. Additionally, the proposed method performs well on datasets with fewer classes, and it can efficiently process multi-attribute datasets.