To study the decision process when decision makers have different risk preferences，a dynamic strategy，which classifies decision makers while approaching the solution，is proposed． First，effective attributes based on the similarity relationship of attribute values are extracted and weights are assigned to them objectively．Second，decision makers’type of risk preferences (risk aversion，risk neutral，risk seeking) are converged into step by step according to the decision activities． According to the possibility and similarity of the interval numbers，the prospect theory models are established． Then，a suitable theoretical model is selected in the process of information aggregation and ranking according to decision makers’risk preferences． Moreover，in order to maximize the fairness of the decision result，we propose a new information fusion algorithm is proposed when different types of decision makers exist at the same time． Finally，the new approach is validated via realistic examples．