In this study, a novel method for the construction of a driving cycle based on Mean Shift clustering is proposed to solve the problems existing in the traditional micro-trips method. Firstly, 1701 kinematic segments are obtained by processing and dividing the driving data in real road conditions. Secondly, 12 kinematic parameters are calculated for each segment, and the dimensionality of parameters is reduced through principal component analysis (PCA). Three principal components are chosen to classify all cycles into three types by the Mean Shift algorithm. Finally, according to the principle of minimum deviation, representative micro-trips are selected from each type of cycle to complete the construction of the final driving cycle. Further, the construction method in this paper is compared with the micro-trips construction method by the K-Means clustering. The results show that the construction method by Mean Shift clustering can more effectively reflect the real driving data. This study realizes the innovation in the construction method of micro-trips and provides a preliminary theoretical basis for the formulation of automobile working condition standards, energy management of new-energy vehicles, and optimal control of vehicle dynamics in driverless vehicles.