Novel Automatic Feature Engineering for Carbon Emissions Prediction Base on Deep Learning

Abstract

The primary cause of global climate change is carbon emissions. The world must urgently reduce carbon emissions to avoid the worst effects of climate change. Understanding the most important features of carbon emissions is the first goal in decreasing carbon emissions. One of the critical issues for carbon emissions is research on feature engineering and prediction. Therefore, we propose a novel automatic feature engineering for carbon emissions. In the proposed algorithm, automatic feature engineering is used to select important features. Furthermore, deep learning is used to reduce the prediction error for carbon emissions. The proposed algorithm, decision trees, and linear regression are compared with previous methods using the Kaggle dataset of carbon emissions. The results demonstrate that the proposed algorithm selects the four most important features from the Kaggle dataset of carbon emissions. The proposed algorithm also enhances and lessens the root mean square error (RMSE) of the prediction. The proposed algorithm outperforms the other approaches.