Application of deep learning models for traffic flow prediction based on time-series data

This paper investigates and applies the Long Short-Term Memory (LSTM) deep learning model for traffic flow prediction based on time-series data. The model is trained and tested with a large dataset comprising 26,497 records of vehicle counts at a specific observation point over four months. To evaluate the performance of the LSTM model, we conduct experiments and compare it with other popular machine learning methods. The results demonstrate that the LSTM deep learning model achieves an accuracy of 88.91%, outperforming traditional machine learning techniques. These results promise to support traffic flow prediction and provide reliable data for managers, helping them make accurate decisions in traffic coordination, thereby reducing congestion and enhancing the efficiency of urban traffic systems.