Layer-2 transferable belief model: Manage uncertainty on random permutation sets

In this paper, the transferable belief model established on power sets is extended to the permutation event space (PES) and is referred to as the layer-2 transferable belief model. Our goal is to provide a comprehensive approach for handling and modeling uncertainty, capable of representing both quantitative and qualitative information. First, the motivation for proposing the layer-2 transferable belief model and its information processing principles are explored from the perspective of weak propensity. Then, based on these principles, the corresponding information processing methods for the credal and pignistic levels are developed. Finally, the advantages of this model are validated through a classifier that leverages attribute fusion to enhance performance and decision-making accuracy.