A Machine Learning for Predictive Traffic Modelling on Nigerian Roads: Bridging Sustainability and Performance

Abstract

The steady rise in vehicle numbers has intensified road congestion, making accurate traffic forecasting essential for effective traffic management in smart city environments, according to SDG 11. This study investigates the application of machine learning (ML) models for real-time traffic flow prediction on Nigerian roads, aiming to overcome the limitations of traditional forecasting techniques, which are often labor-intensive and imprecise. Day-to-day traffic stream forecasting is conducted using three models: The Multilayer Perceptron of Artificial Neural Networks (ANN), the Seasonal Autoregressive Integrated Moving Average (SARIMA), and the Support Vector Machine Regression (SMOreg). The models are trained and tested on six months of real traffic data collected from a road section, with performance evaluated against predefined criteria. Experimental results show that the proposed SMOreg model achieves the highest prediction accuracy, with an R² of 0.9861, 98.61% prediction accuracy, and the lowest absolute relative error of 0.77%. Evaluation metrics, including R-squared (R²), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE) confirm the effectiveness of SMOreg, SARIMA, and MLP, with SMOreg outperforming the others. These findings highlight the potential of ML models to significantly enhance real-time traffic prediction, offering improved accuracy and efficiency over traditional methods.