Cooperative Awareness Message Generation Interval Prediction Model Based on Bayesian Optimized Long Short-Term Memory Neural Network

The prediction accuracy of the generation interval of Cooperative Awareness Message (CAM) is known to directly affect the accuracy of reserved communication resources for CAM in cellular vehicle-to-everything (C-V2X) systems. In order to improve the prediction accuracy of CAM generation interval, a prediction model based on the Bayesian optimized long short-term memory neural network (BAO-LSTM) is proposed. The model utilizes LSTM to predict the generation interval of CAM while employing the Bayesian optimization algorithm iteratively to optimize the hyperparameters of LSTM until they reach their optimal values. The pre-processed Next Generation Simulation (NGSIM) dataset is used to verify the performance of the BAO-LSTM model. Experimental results demonstrate that the BAO-LSTM model outperforms both of the BP model and the LSTM model without the Bayesian optimization process in terms of capturing, and constructing patterns and rules for the generation time interval of CAM.