Abstract

A major factor causing global warming and climate change is carbon emissions. Since the UK intended to attain net zero carbon emissions by 2050, accurate and steady carbon emissions prediction is helpful for developing emissions reduction plans and attaining carbon neutrality as soon as possible Although many prior studies used one or more models to anticipate carbon emissions, there has been little progress in forecasting accuracy and few studies on UK carbon emissions.

Based on UK carbon emissions data from 2005 to 2019, the proposed stacking ensemble model was created by combining three separate base predictive models optimally: random forest, support vector machine and k-nearest neighbour learning models using gradient boosting as meta-regressor. The GridSearchCV optimises the basic model hyperparameters to create a high-performance stacking ensemble model.

The data analysis identifies three high carbon emitting sectors as being the industrial, transportation, and domestic sectors.

The findings indicate that the proposed model accurately forecasted the carbon emissions, with a very low prediction error value. This suggests that the variation between the expected and actual emissions was minimal. The root mean square error for the suggested model was 0.44. Based on the values of the root mean square error, it can be shown that the stacking ensemble model surpasses the random forest- 0.49, gradient boosting- 0.46, k-nearest neighbour- 0.67, and support vector machine- 2.71 models in terms of accuracy. As a result, the proposed model significantly outperforms other models at forecasting, and the results can provide a useful guide for decision-makers as they develop environmental and climate change policies that are linked to the carbon peaking and neutrality targets. However, this study offers a model that is simple to duplicate and use. The model is effective and adaptable to various environmental data types and can also be used to conduct research on other socioeconomic-related issues.

Stacking ensemble machine learning algorithm for predicting carbon emissions and reduction strategies