A method of automatically generating initial parameters for large-scale belief rule base

摘要

The traditional rule-base inference methodology using evidential reasoning approach (RIMER) needs to traverse the reference values of all antecedent attributes when constructing belief rule base (BRB). Therefore, when many attributes have multiple reference levels, the scale of BRB will face the problem of combination explosion. Thus, it is inoperable to require experts to give all the parameters of each rule of BRB model, because of the limitation of expert knowledge in complex problems. If the initial BRB parameters are set unreasonably, the optimization speed and accuracy of the model will be affected. To solve this problem, a method of automatically generating initial parameters for large-scale BRB by using part of the standard rules and cloud model is proposed in this paper. Experts determine the number and parameters of standard rules through prior knowledge based on specific practical problems, and use cloud models to convert qualitative knowledge and quantitative information to automatically generate the remaining rules. A case study is established with Intel Berkeley Research lab data set to verify the effectiveness of the proposed method.