Parameter Learning for an Intuitionistic Fuzzy Belief Rule-Based Systems Based on Weight and Reliability

抄録

The intent of the parameter learning is to ensure the accuracy of intuitionistic fuzzy belief rule-based systems (IFBRBSs) considering both weight and reliability. The main contribution is that distinguish reliability and weight respectively treated as intrinsic and extrinsic properties of evidence. A parameter learning method considering both reliability and weight determined by internal and external conflicts (PL-RW-IEC) is proposed. Evidence reasoning with reliability and weight is introduced as a basis of the learning process. After learning, the mean square error (MSE) between the real output and the simulated output decreases 75 times. Compared to the parameter learning considering both reliability and weight determined by Dempster’s conflict (PL-RW-DC) and compared to the parameter learning not considered reliability (PL-NR), the PL-RW-IEC method gets the most accurate result according to the MSE.