Machine Learning-Based Predictive Model for Fabric Wastage in the Apparel Industry Due to Defective Fabric Raw Materials

In today's competitive apparel industry, maintaining the quality of raw materials is a significant challenge. Defects in fabrics can lead to substantial material and financial losses for manufacturers and retailers. The four-point system, used to identify defective fabrics, accepts batches with fault rates below 28 points per 100 square meters. However, this system still permits some defects, leading to material losses discovered only at the end of the production process. This study aims to improve the prediction of waste losses due to defective raw materials at the fabric inspection stage, thereby reducing material wastage in the apparel industry. Data were collected from six factories of a clothing company in Sri Lanka. Two datasets were used, and after undergoing preprocessing to enhance machine learning performance, one dataset resulted in 633 data points, providing optimal results for material wastage prediction. Ten machine learning models were evaluated, including Multiple Linear Regression (MLR), Support Vector Regressor (SVR), Decision Tree (DT), Random Forest (RF), XGBoost, Genetic Algorithm (GA), Artificial Neural Network (ANN), Convolutional Neural Network (CNN), Simple Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM). Among these, the ANN model yielded the minimum prediction error, with a Mean Absolute Error (MAE) of 22.08. Statistical validation using the Wilcoxon matched-pairs signed-rank test confirmed no significant difference between the predicted and actual values, establishing the ANN model as a reliable tool for minimizing material wastage in the garment industry.