Simplified Extended Belief-Rule-Based System for Classification Problems

Extended Belief Rule-Based (EBRB) system, as an advanced rule-base system, has attracted great attention recently due to its nature of data-knowledge integration, transparency, uncertainty handling, as well as overcoming the limit of combinational explosion of traditional rule-base systems. Despite their strengths, EBRB systems face challenges about computational efficiency and predication accuracy. This study aims to address these challenges by simplifying the EBRB structure through the removal of referential sets of values in implementation, thus enhancing computational efficiency and predication accuracy. Empirical study shows that this approach achieves competitive results compared with benchmarks from other state-of-the-art methods.