Enhancing CO2 emissions prediction for electric vehicles using Greylag Goose Optimization and machine learning

Abstract

Abstract Electric vehicle (EV) $$\hbox {CO}_2$$ emissions should be predicted and mitigated, which requires lowering EV emissions in line with global sustainability goals. Such accurate forecasting supports policymakers and other industry stakeholders make marketing decisions to reduce environmental impacts and optimize resource utilization. In this research, a novel Greylag Goose Optimization (GGO) algorithm is integrated with a Multi-Layer Perceptron (MLP) model to improve $$\hbox {CO}_2$$ emissions prediction. Finally, the study does a comparative analysis with some established optimization algorithms in hyperparameter tuning regarding an improved accuracy model. In addition, statistical analyses such as ANOVA, sensitivity analysis, and T-test were used to substantiate performance differentiation between models. For the optimal model, the GGO-optimized MLP significantly outperformed baseline models and other optimization techniques, having minimum error metrics such as correlation coefficient and RMSE and an MSE of $$4.72 \times 10^{-7}$$ . As a result, the emissions forecast is very reliable. The proposed approach provides actionable insights for environmental policies, EV adoption strategies, and infrastructure planning. The model enables stakeholders to achieve climate objectives, optimize EV charging systems and foster the creation of sustainable transportation systems, as said accurate emissions estimates are enabled.