Search

It is vital to predicting the hidden behavior of complex systems, which is associated with the operation state of the system and cannot be directly observed. In recent research of hidden behavior prediction, one effective method called the semi-quantitative information-based method has received wide applications and become the main stream, where both the qualitative knowledge and the quantitative data are involved. Although some methods have been effectively applied to predict the hidden behavior, only the single behavior prediction has been achieved. It is shown that multiple hidden behaviors are not considered, which may not meet actual requirements of the system. Besides, the influence of perturbations on the system is not considered, which can indirectly affect the system's hidden behavior and reduce the hidden behavior prediction accuracy. As such, a new hidden behavior prediction model based on the evidential reasoning (ER) rule considering perturbations is proposed in this paper, which is typically a semi-quantitative information-based method and can deal with multiple hidden behaviors. A parameter estimation algorithm is proposed based on the expectation maximization (EM) to obtain more accurate model parameters. A case study is conducted to verify the effectiveness of the proposed model. [ABSTRACT FROM AUTHOR]