A modified belief rule based model for uncertain nonlinear systems identification

作者：Ke, Xiaolu; Ma, Liyao; Wang, Yong^*

来源：Journal of Intelligent and Fuzzy Systems, 2017, 32(6): 3879-3891.

DOI：10.3233/IFS-162191

• Belief rule base
• system identification
• evidence theory
• weighted average
• attribute reduction

The belief rule based (BRB) methodology is developed from the traditional IF-THEN rule based system and evidential reasoning (ER) approach. It can be used to model complicated nonlinear causal relationships between antecedent attributes and consequents under different types of uncertainty. In this paper, we present a new BRB structure for modelling uncertain nonlinear systems. It uses the weighted averaging operator to replace the ER approach in the inference process. With this change, the BRB structure could be simplified and faster speeds are obtained in both training and inference process, while universal approximation capability is maintained. By using the consequents of the new BRB model, an approach for reducing possibly redundant referential values of antecedent attributes is proposed for point estimate. Case studies are conducted on three well known benchmark datasets to compare the new model with the existing BRB model and other methods in the literature. Experimental results demonstrate the capability of the proposed method for identification of nonlinear systems.

• 出版日期2017
• 单位中国科学技术大学