Enhancing anomaly detection in IoT-driven factories using Logistic Boosting, Random Forest, and SVM: A comparative...

Area of Science:

• Industrial IoT
• Machine Learning
• Anomaly Detection

Background:

• Industrial environments generate vast sensor data.
• Effective anomaly detection is crucial for operational efficiency and safety.
• Existing machine learning models require evaluation for industrial IoT anomaly detection.

Purpose of the Study:

• To evaluate and compare Logistic Boosting, Random Forest, and Support Vector Machines (SVM) for anomaly detection in industrial IoT.
• To identify the most effective machine learning algorithm for classifying anomalies in factory sensor data.
• To provide insights for developing robust real-time anomaly detection systems.

Main Methods:

• Analysis of a real-world dataset comprising 15,000 instances from factory sensors.
• Utilized Receiver Operating Characteristic (ROC) curves, confusion matrices, and standard performance metrics.
• Comparative evaluation of Logistic Boosting, Random Forest, and Support Vector Machines (SVM) algorithms.

Main Results:

• Logistic Boosting achieved the highest performance with an Area Under the Curve (AUC) of 0.992, 96.6% accuracy, 93.5% precision, 94.8% recall, and an F1-score of 0.941.
• Logistic Boosting demonstrated superior handling of imbalanced data, with 134 false positives and 117 false negatives.
• Random Forest showed strong results (AUC = 0.9982), and SVM exhibited high recall, but Logistic Boosting's ensemble method was most effective.

Conclusions:

• Logistic Boosting is the most effective machine learning algorithm for anomaly detection in industrial IoT environments.
• The findings support the implementation of Logistic Boosting for real-time anomaly detection systems in factories.
• Future research should explore hybrid architectures and edge optimization for enhanced industrial IoT anomaly detection.