Randomness-oriented Multi-dimensional Cloud-based belief rule Base approach for complex system modeling

Abstract

Many complex system modeling problems have to face two major challenges, randomness and interpretability, which are the motives and major contributions of this study. For the randomness challenge, the cloud model is adopted which reveals randomness by assuming the points within an interval following a certain distribution. For the interpretability challenge, the Belief Rule Base (BRB) is adopted which is essentially a white-box with direct access to experts and decision-makers. Therefore, a new approach is proposed to meet the two challenges by combining the cloud model and BRB, namely Cloud-BRB. In the new Cloud-BRB approach, new rule activation and matching degree calculation procedures have been proposed to better accommodate the cloud model into BRB. Moreover, new weight calculation procedures have also been proposed based on the multi-dimensional cloud model which connects the multiple attributes in practical problems and multiple attributes in BRB. A practical case of risk assessment on tunnel-induced damages to pipes is studied. Case study results show the consistency with previous approaches, which validate the efficiency of the proposed Cloud-BRB approach. As the belief distribution and randomness have been included in the Cloud-BRB approach, the results are more informative for decision-makers than previous approaches. Furthermore, Pipe No. 10 has been used to demonstrate the key influential factors identification process by the Cloud-BRB approach.