Comparative Analysis of Machine Learning Algorithms for Automated Hydrophobicity Classification in Composite Insulators Using Contour Features

This study presents a comparative analysis of six machine learning algorithms for automated hydrophobicity classification in composite insulators using contour‐based image features. A robust image‐processing pipeline was developed to extract statistical features from water droplet patterns on insulator surfaces, enabling objective classification of hydrophobicity levels. The models evaluated include gradient boosting (GBT), random forest (RF), support vector machine, k‐nearest neighbours, decision tree (DT) and logistic regression (LR). Performance was assessed using accuracy, precision, recall, F1‐score, receiver operating characteristics (ROC)/area under the curve (AUC) and prediction time. GBT and RF emerged as the most effective classifiers, achieving the highest accuracy (75% and 73%, respectively) and AUC scores (0.94), while maintaining acceptable prediction times. Simpler models such as LR and DT offered faster inference but lower classification reliability. SVM performed the weakest across most metrics. These findings highlight the trade‐offs between accuracy and computational efficiency, with ensemble models offering the best balance for real‐world deployment. The proposed framework demonstrates strong potential for enhancing predictive maintenance in power systems through scalable, noninvasive hydrophobicity monitoring.