In modern manufacturing, preventing defects is essential for cost reduction, quality assurance, and operational efficiency. Traditional quality control methods like Statistical Process Control (SPC) are largely reactive, detecting defects after they occur. With Industry 4.0 and the availability of production data, machine learning (ML) enables predictive quality management. This study compares seven supervised ML algorithms Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, Gradient Boosting Machines, Artificial Neural Network, and k-Nearest Neighbors for defect prediction in a Small and Medium Enterprise (SME). Using a dataset of 3,240 production instances and 17 operational variables, data pre-processing involved normalization, class balancing, and stratified sampling. Models were trained and validated using 10-fold cross-validation and hyper parameter tuning. Evaluation metrics included Accuracy, Precision, Recall, F1-score, ROC-AUC, and confusion matrices. Results showed that ensemble models, particularly Random Forest and Gradient Boosting, outperformed others, achieving high accuracy and balanced error rates, making them ideal for resource-limited SMEs. Decision Tree and Logistic Regression provided interpret-able results with moderate performance, while SVM and ANN achieved high recall but excessive false positives, unsuitable for SMEs with limited inspection capacity. k-NN performed weakest due to instability in high-dimensional data. Overall, ensemble methods offered the best trade-off between predictive accuracy, interpret-ability, and operational efficiency, providing SMEs with a practical framework for proactive defect prevention. The study demonstrates ML’s potential to enhance manufacturing quality control and recommends future exploration of hybrid and IoT-integrated predictive models for real-time quality monitoring.
Chidiebube et al. (Mon,) studied this question.