Enhancing the Prediction of Internet of Vehicles Delay Quality of Service with Clustering Principle

In the communication process of the Internet of Vehicles (IoV), vehicle-mounted units and roadside units are prone to signal interference during information exchange, causing delays that undermine the continuity and real-time property of information interaction and even affect real-time vehicle control and safety decisions. Therefore, predicting the delay performance of IoV services, detecting potential abnormal delays in advance, and taking timely preventive measures are conducive to reducing losses due to delays. To accurately predict the service delay in the IoV, this paper innovatively proposes a random forest training and prediction model enhanced by the clustering principle based on the traditional random forest prediction model. Firstly, data with similar features are clustered into the same group to reduce the computational complexity of random forest training and prediction. Then, through large-scale data training, the prediction accuracy and generalization ability of the traditional random forest model are improved. Experiments show that compared with traditional unclassified and existing classified prediction models, the method proposed in this paper has advantages in terms of root mean square error (RMSE), mean absolute error (MAE), mean bias error (MBE), and coefficient of determination. It provides a novel method for predicting the delay quality of service in the IoV, helps to enhance the reliability and performance of IoV communication, and facilitates the development and application of IoV technology.