Abstract:

The evidence theory is a useful mathematical tool for reasoning with uncertainty and has a wide range of applications in information fusion, pattern recognition, etc. How to construct a proper basic belief assignment (BBA) based on the obtained evidence information, i.e., evidence modeling, is the premise of applying evidence theory. However, the evidence theory does not provide a universal method of evidence modeling. The existing evidence modeling methods are all application oriented. According to the abstract level of available information, the exiting evidence modeling methods are classified into featurebased method and decisionbased method and are summarized respectively. Finally, the problems in evidence modeling are disscussed and the future research directions are predicted.