An XGBoost-Driven Intelligent Classification Model for Textile Product Quality Eligibility: A Case Study at PT ABC Textile

Abstract

Product quality is a critical aspect of the textile industry because it determines whether a product meets the company’s quality standards. This study develops a product eligibility classification model using the XGBoost algorithm to support the Quality Control (QC) process at PT ABC Textile. The novelty of this research lies in positioning XGBoost as an interpretability-driven decision-support tool by integrating real QC inspection data, feature importance and SHAP-based interpretability analysis, and stratified 5-fold cross-validation to support practical QC decision-making. The dataset consists of 500 samples manually labeled based on the company’s quality criteria and includes four technical features: Yarn Strength, Knitting Density, Color, and Defect Level. Data preprocessing involved data cleaning, label transformation, and MinMaxScaler normalization. Model performance was evaluated using stratified 5-fold cross-validation to ensure robust and unbiased assessment. The experimental results demonstrate stable and high classification performance across all folds, with strong class-wise precision, recall, and F1-score values. Confusion matrix analysis indicates that the model performs particularly well in identifying Non-Eligible products, which is critical for minimizing quality risks in industrial applications. Overall, the proposed approach demonstrates that XGBoost can effectively support textile quality control as an interpretable and reliable decision-support system. Future work may explore dataset expansion and cost-sensitive learning to further enhance industrial applicability. Keywords – XGBoost; Classification; Textile Products, Quality Control, Data Mining