Research on intelligent traffic flow forecasting and dynamic dispatching model based on network computing

The study first developed a multi-source data collection and processing system, utilizing microwave radar sensors, weather data, and social media event information, by utilizing distributed storage and computing technologies such as Hadoop and Spark, this approach facilitates the efficient collection, storage, and real-time analysis of traffic data. Building on this foundation, a deep learning model based on Long Short-Term Memory (LSTM) networks was developed, which effectively captures both the temporal dependencies and spatial patterns of traffic flow. Experimental results demonstrate that the model achieves high accuracy in predicting traffic flow, with a Mean Absolute Error (MAE) of 45.3 vehicles per hour, a Root Mean Square Error (RMSE) of 68.7 vehicles per hour, and a Mean Absolute Percentage Error (MAPE) of 12.8%. Furthermore, based on the prediction results, a dynamic dispatching model is designed. By optimizing the traffic signal control and public transportation dispatching strategy, the traffic smoothness and punctuality of public transportation services are significantly improved. The average delay time of vehicles in peak hours is reduced by 37.2%, the traffic efficiency of road network is improved by 19.8%, and the punctuality rate of public transportation is improved by 19.8%. The application of this research in large-scale traffic flow forecasting and dynamic dispatching shows good performance and scalability, this offers fresh perspectives and approaches for advancing intelligent transportation systems (ITS).