To increase the dispersion of the objective weights derived from entropy weighting method, an improved approach is presented for normalizing the decision matrix based on adjustment coefficient. The entropy of attribute evaluated values is used as a measurement of data stability, and the objective weights of attributes are determined through their entropies according to the positive relationship between data stability and attribute weights based on the same alternative and attribute but different decision-makers. Subjective weights of attributes are obtained on the basis of the stability of subjective preference values given by group decision-makers and average of them. Finally, an example is given to verify the feasibility and rationality of the proposed method.