Urban road operating speed prediction model based on IPSO-BiLSTM

Existing vehicle speed prediction research usually focuses on expressways or highways. A prediction method based on IPSO-BiLSTM is proposed for operating vehicle speeds on all levels of urban roads in this paper. First, the structure, parameter setting and training process of the multi-layer BiLSTM model were analyzed. The time series data of road operating speed and traffic congestion index constituted the input of the prediction model to establish the operating speed prediction model. Secondly, particle swarm optimization algorithm was studied and random distribution is used to optimize its weight. Then, the improved algorithm is used to optimize the number of hidden layer nodes, learning rate and number of iterations of the prediction model. Finally, the LSTM model and the GRU model were introduced to make predictions using actual road operating data with different road conditions in Guangzhou to compare the prediction performance of each. The results show that IPSO-BiLSTM model has significantly improved prediction accuracy, compared with LSTM model and GRU model.