GPA: Uni-directional GRU-Based Traffic Prediction Model for Minimizing Air Pollution

In today’s world, tackling air pollution is a significant challenge, and the transportation system stands out as one of the main culprits behind this issue. Consequently, this paper focuses on developing a future traffic flow forecasting system. To accomplish this, we used a Gated Recurrent Unit (GRU) and a large dataset of time-series data to construct the GPA model, which aims to predict future traffic flow. The GPA model effectively captures the long-term patterns within the sequential big dataset and incorporates traffic flow characteristics from various junctions. The general public needs an accurate and timely predictor of future traffic flow to help mitigate air pollution. To address this demand, this paper employs time-series analysis and explores the spatial and temporal correlations within traffic data for traffic prediction. The evaluation results demonstrate the performance of the GPA model, indicating that it outperforms the autoregressive integrated moving average (ARIMA) and seasonal autoregressive integrated moving average (SARIMA) models by achieving a 40% and 50% improvement in future traffic flow prediction, respectively, through network loss minimization. The network loss is measured using the root-mean-squared error, while model performance is assessed using the mean absolute error and mean squared error metrics.