MUH: Maximum-uncertainty-heuristic method of modeling belief function

Abstract

Modeling of belief functions is the basic step for using belief function theory to deal with uncertainty. Transformation method for belief functions from different information representation has always been an important method of building belief function, which can integrate multi-source information and vary diverse modeling approaches. Additionally, building belief functions with constraints under maximum uncertainty is a promising way due to various forms of uncertainty measures. However, their drawbacks such as low flexibility and high computational complexity restrict their widespread application. To address these issues, a normalized dynamic total uncertainty measure is proposed, which has a simple and flexible form to obtain belief functions with constraints. Then, a heuristic method is proposed to solve the problem of high computational complexity. The proposed method is validated based on numerous experimental results compared to other methods.