Aiming at the objective evaluation of shift quality of DCT (dual clutch transmission) vehicles, 10 parameters are selected as the evaluation indexes of shifting quality from two evaluation dimensions of smoothness and power performance. In order to solve the problems of redundancy between indicators and the lack of rationality of the existing weighting methods, based on the primary evaluation indicators, the evaluation indicators are reduced by using the rough set knowledge reduction method optimized by genetic algorithm, and the effectiveness of the reduction is verified by using the classification model of support vector machine. Then, based on the concept of attribute importance, the reduced evaluation index is weighted. The test results show that the method can effectively delete redundant indicators, and reasonably weight the evaluation indexes, which provides a basis for objective evaluation of shift quality of DCT vehicles.