Forecasting CO2 Emissions Using Machine Learning And Deep Learning Models: A Comparative Study Across Techniques

Abstract

This study investigates the application of modern machine learning (ML) and deep learning (DL) models to predict carbon-dioxide (CO2) emissions across 93 countries over the period 1970–2030. By utilizing historical CO2 emissions data, we trained various predictive models using data from 1970 to 2015 and evaluated their performance on data from 2016 to 2023. Forecasts were then made up to the year 2030. Eight models were implemented: Artificial Neural Network (ANN), Bayesian Neural Network (BNN), Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), Deep Neural Network (DNN), Long Short-Term Memory (LSTM), Random Forest (RF), and Gra dient Boosting Machine (GBM). Results indicate that the LSTM and BNN models performed best in capturing temporal patterns and uncertainty. The study highlights how accurate models can offer valuable insights for climate policy and emission reduction strategies. Accurate CO2 emission fore casts can help countries plan for sustainable development by balancing growth and environmental care. This research supports long-term sustainability by providing data that can guide smart and climate-friendly decisions.