Abstract:The problem of excavating knowledge from historical parking data and forecasting the number of parking spaces in a short period is studied.By analyzing the factors that affect parking space, we establish a BP neural network in which the network input variables are defined through the combination of time series.Then, a self-adaptive studying rate is used in different stage of training and the momentum terms are added to improve the convergence of the network.According to the real data collected from a large underground parking in town, the simulation and analysis are executed based on Matlab, which results in well-accepted prediction effect.The conclusion shows that the proposed method can improve the prediction accuracy compared with the traditional time series prediction method.