A Comparative Analysis of Machine Learning Algorithms for Detecting Spindle Anomaly in DTY Machine: Optimized Random Forest vs SVM

Abstract This study presents a comparative analysis of two machine learning algorithms—Random Forest (RF) and Support Vector Machine (SVM)—for detecting spindle anomalies using vibration signal features. A structured workflow was implemented that involved signal acquisition, feature extraction, preprocessing, and model training. To enhance classification accuracy, both models were optimized using the Randomized Search CV method with 5-fold cross-validation. Evaluation metrics included accuracy, precision, recall, F1-score, confusion matrix, and the Receiver Operating Characteristic (ROC) curve analysis. Before hyperparameter tuning, RF demonstrated superior performance compared to SVM, particularly in abnormal condition detection. After the optimization process, SVM exhibited substantial improvements across all metrics, achieving an accuracy of 96% and a perfect Area Under the Curve (AUC) score of 1.00. In comparison, RF maintained stable, balanced performance, with an accuracy of 94% and an AUC of 0.99. Training time analysis further revealed that SVM is significantly faster, making it more suitable for real-time applications. These results highlight the effectiveness of model tuning in improving fault detection and demonstrate the potential of both RF and SVM as reliable tools for predictive maintenance in DTY machines. Moreover, the findings offer practical insights for implementing machine-learning-based solutions in industrial vibration monitoring, particularly in the textile manufacturing sector.