Iov Traffic Prediction Utilizing Bidirectional Memory And Spatiotemporal Constraints With Local Search And Non-Linear Analysis

Abstract

The Internet of Vehicles (IoV) is fast expanding depending on accurate traffic prediction to maximize travel paths, increase road safety, and reduce congestion. Traditional traffic prediction systems produce less than perfect performance in dynamic environments since they cannot capture the complex spatiotemporal dependencies and non-linear traffic patterns inherent in IoV networks. This work addresses these challenges by way of an upgraded traffic prediction model combining Bidirectional Long Short-Term Memory (Bi-LSTM) networks with spatiotemporal restrictions and a local search optimization method. The model uses Bi-LSTM to efficiently capture the temporal dependencies from past and future traffic data, while the spatiotemporal constraints boost the model power to grasp spatial correlations among surrounding road segments. The model parameters are tuned using a local search technique, and non-linear analysis is applied to identify and modify traffic flow abnormalities, thereby improving the prediction accuracy. The proposed approach shown superior performance than more conventional approaches on a big-scale IoV traffic dataset. In MAE, specifically, the model exceeded earlier methods by 12.6%; in RMSE, by 15.4%; and in MAPE, by 10.8%. Its Root Mean Square Error (RMSE) was 6.89 while its Mean Absolute Error (MAE) was 4.57. These results indicate the adaptability of the model since they illustrate how well it catches the dynamic and complex character of IoV traffic.