Research on Parking Space Prediction Based on Long Short-Term Memory

The berth occupancy prediction is a key issue in the intelligent parking guidance information system. Current neural network-based prediction techniques will dramatically decrease in prediction accuracy as the prediction step size increases. In order to solve the problem that the prediction accuracy of berth occupancy decreases with the increase of step size, we propose a time series prediction model based on long short-term memory (LSTM) neural network. The model enhances the memory function of the network by controlling the forget gate, output gate and input gate, filtering out relatively unimportant information and reducing the sample dependence of the model, thus it improves the training efficiency. The model is tested by using real parking data. The test results show that the proposed model has certain advantages in the accuracy of long-term prediction.