Comparative Study of Machine Learning for Predicting Compressive Strength in Oyster Shell Cementitious Composites.

Area of Science:

• Materials Science
• Environmental Engineering
• Civil Engineering

Background:

• Annual oyster production generates substantial waste oyster shells, leading to environmental issues due to improper disposal.
• Developing sustainable construction materials from waste products is crucial for environmental protection and resource management.

Purpose of the Study:

• To predict the compressive strength of cementitious composites incorporating waste oyster shells using machine learning.
• To identify the most influential factors affecting the compressive strength of these composites.
• To evaluate the potential of waste oyster shells as a substitute material in construction.

Main Methods:

• Collected and utilized 336 datasets, including experimental results and literature data.
• Compared various machine learning algorithms: Ridge Regression, Support Vector Regression, Artificial Neural Network, and Random Forest.
• Optimized ensemble algorithms (XGBoost, AdaBoost, Extra Trees, LightGBM) using GridSearchCV.

Main Results:

• Random Forest achieved high predictive performance (R² = 0.8411).
• LightGBM demonstrated superior predictive capability with R² = 0.9042, MAE = 3.1671, MSE = 17.8054, and RMSE = 4.2196.
• SHAP analysis identified water-to-cement ratio (W/C) and superplasticizer content as key variables. Oyster shells showed a negative correlation with sand, suggesting their role as a partial replacement.

Conclusions:

• Waste oyster shells can be effectively incorporated into cementitious composites to maintain compressive strength.
• Machine learning, especially LightGBM, provides accurate predictions for the performance of these sustainable materials.
• The findings support the use of oyster shell composites as environmentally friendly construction materials with appropriate admixture optimization.