Large-Scale linguistic Z-Number Belief Rule Base Methodology for Multidimensional and Unreliable Knowledge Representation and Learning

With excellent interpretability, the fuzzy rule-based method stands as a formidable instrument for knowledge representation and learning. Nowadays, the knowledge representation problem with multidimensional input information is widespread, leading to a large rule base and making it difficult to embed expert knowledge. In addition, human knowledge is not entirely reliable, causing inaccurate reasoning results. In this article, a novel large-scale linguistic Z-number belief rule base (LSLZ-BRB) method is proposed for the above multidimensional and unreliable knowledge representation and learning. Specifically, a multidimensional knowledge mapping representation method under the probabilistic framework is proposed to generate an LSLZ-BRB. It allows experts to embed knowledge via conditional probability and prior probability. To reduce the modeling error caused by uncertainty of knowledge, an online interactive learning mechanism of uncertain knowledge is developed. This mechanism ensures that LSLZ-BRB has high real-time performance and improves the accuracy of knowledge representation. A performance evaluation case for the laser inertial measurement unit (LIMU) and experiments on some public datasets illustrate the implementation process of the proposed method and further verify its effectiveness.