Enhancing anomaly detection in IoT-driven factories using Logistic Boosting, Random Forest, and SVM: A comparative machine learning approach
Mohammed Aly, Mohamed H. Behiry
Abstract
Three machine learning algorithms-Logistic Boosting, Random Forest, and Support Vector Machines (SVM)-were evaluated for anomaly detection in IoT-driven industrial environments. A real-world dataset of 15,000 instances from factory sensors was analyzed using ROC curves, confusion matrices, and standard metrics. Logistic Boosting outperformed other models with an AUC of 0.992 (96.6% accuracy, 93.5% precision, 94.8% recall, F1-score = 0.941), demonstrating superior handling of imbalanced data (134 FPs, 117 FNs). While Random Forest achieved strong results (AUC = 0.982) and SVM showed high recall, Logistic Boosting's ensemble approach proved most effective for industrial IoT classification. The findings provide actionable insights for real-time detection systems and suggest future directions in hybrid architectures and edge optimization.