An efficient algorithm for identification of real belief measures

Belief measures are widely applied to management of uncertainty in information fusion. In most published applications, the estimations of belief measures that come from empirical rescouses, such as expert systems, are considered to be real belief measures without any validation. We proposed an efficient algorithm that can quickly detect the contradiction between the estimation and requirements of a real belief measure and adjust the estimation accordingly. The contradiction is assessed by a probability assignment and the estimation is adjusted by Genetic Algorithm. We tested the algorithm using two different simulations. As a result, it shows that the proposed algorithm successfully identified the real belief measures.