A Novel Stacked Machine Learning Framework for Enhanced Prediction of Corporate CO₂ Emissions in Multisectoral Analysis

Climate change and global warming are among the most pressing issues of our time, drawing worldwide attention due to their observable effects such as extreme weather events, heatwaves, flooding, and droughts. The major driver of these phenomena is the excessive carbon dioxide $\left(\mathbf{C O}_{2}\right)$ emissions into the atmosphere, where corporations play a crucial role. This research aims to accurately predict CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emissions in the corporate sector. By providing precise predictions, corporations can gain valuable insights to inform decision-making and strategically plan their carbon neutrality initiatives in alignment with sustainability goals. The data used for this research was collected from 17 worldwide companies from 2017 to 2022. This study introduces a novel stacked model that leverages the strengths of four base models: Ridge Regression, XGBoost, CatBoost, and LSTM. A meta-learner, developed using a Random Forest algorithm, aggregates the predictions from these base models. The introduced stacked model significantly outperforms each base model, achieving the lowest MAE and RMSE values and the highest $\mathbf{R}^{\mathbf{2}}$ of 0.9587, demonstrating superior predictive power. This research contributes to corporate sustainability efforts and offers a wider societal benefit by encouraging reductions in CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emissions, ultimately supporting the fight against climate change and protecting the entire ecosystem.