A multi-stage information aggregation method is studied, which concerns the influences of both decision information and judgment information. Specifically, an analysis frame is established according to the dual information structure which includes decision support information and experts’ preference information. Additionally, a programming is constructed in multi-stage, in which the objective is minimizing the differences of the above decision information. Furthermore, a preference degree is studied to aggregate the information of both dynamic comprehensive preferences and expert’s judgment information. Finally, a numerical case is conducted to verify the feasibility and rationality of the proposed method.