The Dempster-Shafer theory of evidence is currently one of the main mathematical frameworks used for reasoning under uncertainty, as it allows combination of imprecise information originating from different sources. In multi-source information fusion it may be problematic to identify which of the multiple information sources had the most significant contribution to the final decision. This may be of particularly importance in situations where it may not be possible to accurately assess the reliability of the sources involved. As the results of information fusion may be used in safety critical situations it is necessary to be able to track the origin of the beliefs obtained. In order to address this question we propose a new measure based on Jousselme's distance between two basic probability assignments. The concept of measuring the dissimilarity between two bodies of evidence is extended by conditioning the final fusion outcome on the plausibility of the proposition in question. The assumption is that the more similar a body of evidence provided is to the plausibility-conditioned result, the more significant its contribution. The performance of this metric is experimentally tested in the context of identifying the main source of information contributing to a hypothesis and sensitivity analysis.
|Title of host publication||2017 13th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery (ICNC-FSKD 2017)|
|Subtitle of host publication||Proceedings of a meeting held 29-31 July 2017, Guilin, China|
|Publisher||Institute of Electrical and Electronics Engineers (IEEE)|
|Number of pages||8|
|Publication status||Published - Jul 2018|