Parking difficulties and congestion are becoming more and more serious. Informing drivers in advance of the number of empty parking spaces in the future can reduce their time to find effective parking spaces, so as to alleviate the congestion. Based on this phenomenon, an empty parking space prediction model based on LightGBM-SVR-LSTM was proposed. Firstly, through data processing, some noise data were repaired on the basis of preserving the characteristics of the original data as much as possible. Secondly, the repaired data were put into the light gradient boosting machine (LightGBM), the value of leaf node were extracted as new features, and were put into the support vector regression (SVR) for prediction. Then, the long short-term memory neural network (LSTM) was used to repair the error. Finally, the data of a parking area were selected to verify the prediction effect by using root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE). The results show that the accuracy of the proposed combined model is improved and has robustness under normal conditions and holidays.