Based on the arrival-departure time data of vehicles in stereo garage, k-means clustering method was used to classify vehicles according to the arrival frequency of access vehicles in different periods, and cubic cluster criterion was used as the evaluation index to evaluate the classification credibility. Based on the reasoning results of vehicle arrival-departure time division and the relationship between the total service time of equipment from I/O to the parking space and the length of stay time, a mathematical model of parking space partition allocation in stereo garage is established. With defining the average customer waiting time as stereo garage efficiency evaluation index, the efficiency index simulations of the nearby allocation and the proposed partition clustering reasoning allocation were carried out. The simulation results show that the proposed allocation strategy, compared with nearby allocation strategy, can effectively shorten the customer waiting time, and the customer waiting time reduced by 9.5%. The results provide reference for the parking space allocation process of such garages, and provide decision support for improving the operation efficiency of garages.