Enhancing clothing industry quality: predicting knitwear defects with data mining

Purpose In the apparel industry, companies aim to remain competitive by producing high-quality fashion products at low production costs. This drive to balance quality and cost directly affects the business’s profitability. Therefore, developing a robust quality assurance process is essential for clothing manufacturers. This study aims to design a model that predicts potential defect types based on the technical requirements of clothing orders, using data collection, analysis and product quality management knowledge. Design/methodology/approach In quality assurance, understanding the reasons for the forecast results and, from there, having a basis for the reverse effect on the production line is one of the most critical requirements to improve product quality. In this research, the decision tree and Naïve Bayesian technique – two kinds of white box models were applied to develop the forecasting methodology of defect types in apparel order based on input material and sewing technical parameters. Findings This research identified four apparel-order factors influencing product quality: fiber composition, fabric structure, fabric weight and stitch density. Four factors were applied through data collection in historical data and data standardization–normalization process to propose a forecasting model. Defect forecasting results from the decision tree had higher correct classification instances than the Naïve Bayesian technique. Originality/value This research provides new insight into the relationship between the characteristics and input technical requirements of clothing orders on final garment quality. Combining a precise process and scientific knowledge could improve the efficiency of quality assurance in the clothing industry, especially in knitting garments.