Considering the mixed form of attribute evaluation values and the uncertainty of language evaluation by decision makers, this paper proposes a hybrid multi-attribute group decision-making method which considers the weights and reliabilities of evidence. First, the mixed form of attribute evaluation values is transformed into unified interval belief structures. Then, the interval belief structures of quantitative and qualitative indicators of alternatives are fused by using the nonlinear optimization models based on evidential reasoning (ER) rule, and the distributed evaluation vectors of the total interval belief structures of the alternatives under the information fusion strategies of ``attributes synthesis before expert decision-making'' and ``expert decision-making before attributes synthesis'' are obtained, and the alternatives are ranked according to the expected utility theory. Finally, a supplier selection case is used to verify the feasibility and effectiveness of the proposed method. The analysis results show that the proposed method in this paper simultaneously considers the weight and reliability of the evidence and the heterogeneity of the attribute evaluation information. It not only conforms to the realistic decision-making, but also has better ranking consistency and information integrity in the information fusion process compared with the existing method, the decision method proposed in this article can be regarded as an effective extension of the existing method.