Recently, the phenomenon of large parking gaps and low utilization of parking space is common. Based on the analysis of shared berth demand and supply, this study establishes a behavior model of shared berth selection. According to a two-day field survey on the occupied parking spaces, the number of vehicles arriving and leaving the West courtyard of Lijiacun, this paper analyzes the personal attributes, daily parking attributes, sharing attributes, shared security, and economic attributes, and convenient and flexible perception attributes of parking users. Then, the behavior model of berth-sharing selection is established by using the SEM-logit combination model, and the characteristics of travelers' sharing demand are analyzed. In the aspect of shared parking space supply, BP neural network and SVM are used to predict the supply capacity of regional parking lots. The results show that SVM has higher precision prediction and can provide data support for shared parking service platforms.