Decision Integrated Tree: A Machine Learning Model to Predict CO₂ Emission from Power Plants

Carbon dioxide (CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>) emissions stem from human activities like burning fossil fuels, contributing to the greenhouse effect and climate change. In the United States, the electricity generation sector is a significant emitter, primarily from coal and natural gas. Efforts to transition to cleaner energy sources aim to mitigate CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emissions and their adverse environmental impacts (global warming, sea level rise, and disruptions to ecosystems and biodiversity). In this research, a novel approach called Decision Integrated Tree (DIT) is presented, which investigates with a large dataset, captures complex relationships, addresses non-linearity and complexity in CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emission data. It commences with dataset collection (containing 5094 data points and six features), followed by thorough data preparation involving cleaning, preprocessing, and feature engineering to ensure data quality. Subsequently, our proposed DIT model is applied, incorporating various techniques (ensemble methods, pruning, and regularization). Experiments evaluate the performance of the model using significant metrics including R-squared Score ($\mathbf{R}^{2}$), Mean Squared Error (MSE), and Mean Absolute Error (MAE). From the experiment, we obtained $\mathbf{9 8. 2 \%} \mathrm{R}^{\mathbf{2}}, \mathbf{0. 6 3 4}$ MSE, and 0.421 MAE predicting CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emissions from power plants. The integration of ensemble methods, feature engineering, and hybrid models within the DIT framework enhances predictive capabilities and informs decision-making in environmental policy, industry, and research. Strong performance metrics validate the model’s reliability, while a variegated large dataset enable robust predictions in CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emission forecasting. However, deep learning integration and more data inclusion are needed to enhance the performance of the model and applicability for future advancements in environmental modeling.