Responsible AI: A Comparative Analysis of Different Machine Learning Classification Models on Energy Consumption and Classification Accuracy

Artificial Intelligence (AI) and Machine Learning (ML) have significantly transformed various industries, including healthcare, finance, autonomous systems, and environmental monitoring. These technologies enable advanced data processing, predictive analytics, and decision-making capabilities, driving efficiency and innovation. However, the increasing reliance on ML models, especially deep learning and ensemble techniques, has led to growing concerns regarding computational cost, energy consumption, and environmental impact. The carbon footprint associated with large-scale model training is substantial, requiring researchers and developers to explore energy-efficient AI methodologies. This paper systematically evaluates classification algorithms, including Decision Trees, Naive Bayes, Random Forest, Support Vector Machines (SVM), and Gradient Boosting, in terms of computational efficiency, accuracy, training time, and carbon emissions. By leveraging the CodeCarbon library, we measure the energy consumption and CO₂ emissions of each model to assess their sustainability impact. The study aims to highlight the trade-offs between model complexity, performance, and environmental footprint, providing insights into how algorithm selection affects both computational efficiency and sustainability. The findings can guide AI practitioners in developing more environmentally responsible models while maintaining high accuracy and predictive power.