A Comparative Study of Machine Learning Algorithms for Anomaly Detection in Industrial Environments: Performance and Environmental Impact

This research explores the intersection of artificial intelligence and environmental sustainability, specifically in the context of anomaly detection in Industry 4.0. The study evaluates a range of machine learning algorithms and various configurations of Multi-layer Perceptron (MLP), using several metrics to assess model performance alongside their environmental footprint. Findings show that traditional algorithms such as Decision Trees and Random Forests offer reliable efficiency and performance. In contrast, optimized MLP configurations can yield superior results, albeit with greater resource consumption. The research underscores the importance of balancing performance, complexity, and environmental impact in model development. Through a multi-objective optimization approach based on Pareto optimality principles, it reveals trade-offs between model performance and environmental considerations. The study thus serves as a guide for creating environmentally responsible machine learning models for industrial applications.