Uddin, Md NasirShi, Xijun2024-04-222024-04-222024-04-02Uddin, M. N., & Shi, X. (2024). Advancing lightweight engineered cementitious composites: An interpretable machine learning framework. Poster presented at the Graduate Student Research Conference, San Marcos, Texas.https://hdl.handle.net/10877/18490In recent years, the integration of Machine Learning (ML) techniques to predict the properties of Lightweight Engineered Cementitious Composites (LWECCs) has garnered significant attention. The Compressive Strength (CS) and Flexural Strength (FS) are pivotal attributes of LWECCs, underpinning their utility in various civil engineering endeavors. This research aims to collate mixture design components and their associated strengths of LWECCs, specifically those reinforced with polyethylene, and polyvinyl alcohol fibers, from the extant literature. To predict the CS and FS of LWECCs, models based on eXtreme Gradient Boosting (XGBoost) and Light Gradient Boosting Machine (LightGBM) were developed. Emphasis was placed on hyperparameter optimization using GridSearchCV to refine model performance for LWECCs. Additionally, the influence of mixture properties on model outcomes was investigated through SHapley Additive exPlanations (SHAP) analysis, providing insights into optimal mixture designs for LWECCs. This study underscores the potential of enhancing predictive modeling in civil engineering by integrating advancements in machine learning, offering a pathway to more effective and efficient material design.Image1 page1 file (.pdf)enCC0 1.0 Universalmachine learninglightweight engineered cementitious compositesmechanical strengthoptimizationPoster