Enterprise AI Analysis

Enhancing Dental Polymer Formulation through Interpretable Machine Learning: A Comparative Analysis of Feature Selection and Algorithm Performance

The development of high-performance dental polymer composites remains a significant challenge in restorative dentistry, requiring simultaneous optimization of mechanical strength, dimensional stability, and optical properties. This study presents a comprehensive comparative analysis of machine learning algorithms for predicting dental polymer performance, coupled with advanced feature selection methodologies. We systematically evaluate Random Forest, XGBoost, LightGBM, Support Vector Regression, and Gaussian Process Regression on a curated dataset of 347 original dental composite formulations (expanded to 390 training samples through data augmentation applied exclusively to the training set). Using a hybrid feature selection approach combining filter, wrapper, and embedded methods, we reduce the feature space from 283 to 67 critical descriptors while achieving R² values of 0.923 for flexural strength and 0.897 for polymerization shrinkage predictions. SHAP-based interpretability analysis reveals that resin monomer ratio, filler loading percentage, and particle size distribution constitute the dominant factors governing mechanical performance. The framework is estimated to accelerate formulation design cycles by approximately 67% compared to traditional experimental approaches, based on scenario-based analyses using typical industrial baselines.