Predictive Modelling for Sustainable Traffic Management

Enhancing the accuracy in predicting energy needs involves combining predictive modelling with the integration of traffic flow, accidents, and congestion data. Given the real-time nature of traffic data, accurate energy forecasting is crucial for smart grid management. This study compares Linear Regression, Random Forest, Decision Tree, Support Vector Regressor (SVR), and LSTM neural networks, along with a Gradient Boosting Regressor (GBR) meta-learner ensemble. Using multi-source data (traffic volume, energy consumption, and weather), the ensemble model achieved the best performance with R² = 0.9907, MAE = 162.6 kWh, RMSE = 213.3 kWh, and MAPE = 3.2%. The work supports sustainable urban energy management aligned with SDG-7.