Research on the Fusion of Dependent Evidence Based on Mutual Information
Xiaoyan Su, Lusu Li, Fengjian Shi, Hong Qian · IEEE Access · 2018
The Dempster-Shafer evidence theory has been investigated for many applications due to its ability in handling uncertainty and ignorance. However, the classical Dempster's combination rule can only be applied to the cases, where evidence is independent. This assumption is often unrealistic and may lead to unreasonable decisions. In this paper, a new method for combining dependent evidence based on mutual information is proposed. First, the mutual information is used to measure the dependence degree between evidence. Second, the total discount coefficient is defined based on the dependence degree between evidence. Finally, the aggregation model based on the total discount coefficient and Dempster's combination rule is presented in the information fusion stage. The experiments on Iris data is illustrated to show the use and effectiveness of the proposed method. Compared with other methods, the proposed model has the highest classification recognition accuracy.