In many cases, decision making is to order alternatives and select one or a few among the top of the ranking based on decision makers' preferences. Due to the complexity and flexibility of reality, decision makers usually can only provide partially ordered preferences with certain belief degrees. In order to deal with this kind of decision making problem, we introduce in this paper the use a belief structure for representing the involved partially ordered preferences with belief degrees, then a preference combination approach based on evidential reasoning is proposed to combine the preferences from different decision makers. The final ranking of alternatives is generated based on a distance measure between pairs of preference relations. An illustrative example is provided to show the rationality of the proposed method.