For preference reversal frequently occurring in the multiple criteria decision making(MCDM), a modified MCDM model based on the TOPSIS method is proposed, in which the vector normalization method is replaced by the MAX method and the similarity distance of alternatives is used to measure the advantages and disadvantages of each option. This method is not only simple to calculate, but also can better measure the differences among alternatives, making the decision result more accurate. Comparing the results of the modified model with those of SAW, AHP, TOPSIS and VIKOR methods, it is found that when there are only original options, the ranking results of the model proposed are consistent with those of SAW, AHP methods. When an option is added or deleted, other methods will produce different degrees of preference reversal, but the modified model can keep the relative rank of alternatives, showing the proposed model is more effective and reliable in avoiding preference reversal compared with SAW, AHP, TOPSIS and VIKOR methods.