Neural networks in environmental science: Forecasting CO2 emissions

Abstract

The aim of the study is to analyze the performance of neural network models in predicting CO2 emissions across various regions, including major emitters such as China, the USA, the European Union, India, and Japan. Utilizing a dataset of carbon dioxide (CO2) emissions in the industrial sector from 2000 to 2022, the study employs three neural network models (LSTM); (RNN); (MLP); using Python for analysis. The results indicate that RNN consistently outperformed the other models, achieving the lowest mean square error (MSE) and root mean square error (RMSE). Country-specific analyses revealed significant challenges in India, where all models struggled, while RNN excelled in China and the United States. In contrast, Japan's LSTM model underperformed. Overall, CO2 emissions are projected to decline in India and China, driven by environmental policies and technological innovations, whereas emissions in the United States show only modest reductions. These findings highlight the importance of tailored forecasting approaches and effective environmental strategies for achieving sustainability.