A COMPARATIVE ANALYSIS OF MACHINE LEARNING ALGORITHMS FOR PREDICTIVE MAINTENANCE IN INDUSTRIAL SYSTEM

Abstract In modern industrial systems, equipment downtime and unexpected failures can result in significant production losses, high maintenance costs, and safety risks. Predictive maintenance (PdM) has emerged as a vital solution to these challenges by leveraging machine learning algorithms to forecast potential machine failures before they occur. This study presents a comparative analysis of three supervised machine learning algorithmsSupport Vector Machine (SVM), Random Forest (RF), and Gradient Boosting (GB)for predicting machine failures in industrial environments. The model implementation was carried out using the AI4I 2020 Predictive Maintenance dataset, while the system was developed in Python with a Streamlit-based user interface to enhance accessibility and real-time analysis. The preprocessing phase involved handling missing values, scaling features, and addressing class imbalance through Synthetic Minority Oversampling Technique (SMOTE) to improve model generalization. Evaluation metrics such as accuracy, precision, recall, F1-score, confusion matrix, and ROC-AUC curve were used to assess model performance. Experimental results revealed that all three models achieved high predictive accuracy, with Random Forest demonstrating slightly superior overall performance in terms of precision and recall. The developed system allows users to flexibly upload datasets, select desired algorithms for analysis, and visualize results interactively. The findings of this research highlight the effectiveness of ensemble-based algorithms for predictive maintenance and underscore the importance of data balancing and proper feature scaling in enhancing fault prediction accuracy.