Towards Smart Cities: Practical Swarm Optimization and Long Short-Term Memory for Short-Term Traffic Prediction and Management in Oman

Abstract

Traffic congestion forecasting is a serious challenge to Omani cities, especially Muscat, due to population growth, rapid urbanization, and heavy reliance on private vehicles. Despite government efforts to improve infrastructure, the congestion problem persists. Furthermore, few studies have addressed the issue of congestion in Oman, as this is due to the lack of a national dataset that can be used for prediction studies. Therefore, this study aims to generate a notional dataset and propose a hybrid model that combines a deep learning model with an optimization algorithm to provide accurate predictions for the short term. This work is expected to contribute to bridging an important knowledge gap and establishing a scientific basis for developing smart transportation applications in Oman. Furthermore, in the future, the model can be enhanced to use IoT techniques to collect data and extend the model to predict for the long term on multiple roads. The proposed hybrid PSO-LSTM model provides accurate predictions, with MAE of 1.7185 vehicles, an RMSE of 1.3109 vehicles, R2 of 0.9460 and an MAPE of 1.75%. It provides superior performance compared to the standalone LSTM.