“…Due to its superiority on manipulating the uncertain information, it was diversely applied into machine learning [3], like rough set [6], [7], fuzzy set [4], [5], Z value [8], [9], D number [11], [12], belief structure [10], soft likelihood function (SLF) [13], confidence function [18] and belief entropy [16], [17]. As a powerful tool for analyzing the fusion and expression of decision-level uncertainty information, D-S evidence theory appears to offer sufficiently broad applicability in fault diagnosis [21], [9], decision analysis [22], [23], and information fusion [19], [20]. Particularly, the evidence information fusion from multiple independent sources is mainly performed with Dempster's combination rule.…”