<![CDATA[This study develops an interpretable Random Forest workflow to predict silk fabric quality from simulated physical material feature data. The model uses 150 controlled observations with eight predictors: fabric thickness, fabric weight, thread density, tensile strength, elasticity, luster, softness, and surface defects. The simulated values were generated with class-specific truncated normal distributions based on measurable textile properties, standard testing variables, and domain assumptions. Data were evaluated using a stratified 80:20 holdout, stratified 5-fold cross-validation, weighted accuracy-related metrics, and confusion matrix analysis. Random Forest achieved 0.87 holdout accuracy, 0.88 precision, 0.87 recall, and 0.87 F1-score. It showed competitive performance against KNN, Decision Tree, SVM, ANN, and XGBoost while providing direct feature-importance interpretation. Surface defects, tensile strength, and thread density were the most influential predictors. The main contribution is a reproducible simulation and interpretation framework for early-stage silk fabric quality prediction. Because the dataset is simulated, the results demonstrate method feasibility and require validation with real textile production or laboratory data.]]>
Copyrights © 2026
