Use of BOIvy Optimization Algorithm-Based Machine Learning Models in Predicting the Compressive Strength of Bentonite...

Area of Science:

• Materials Science and Engineering
• Civil Engineering
• Artificial Intelligence in Materials

Background:

• Bentonite plastic concrete (BPC) offers structural and heavy metal adsorption benefits.
• Accurate compressive strength (CS) prediction is vital for BPC design.
• Traditional CS testing methods are time-consuming, costly, and uncertain.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for predicting BPC compressive strength (CS).
• To enhance ML model prediction accuracy using meta-heuristic optimization.
• To identify key factors influencing BPC CS.

Main Methods:

• Machine learning models including Support Vector Regression (SVR), Artificial Neural Network (ANN), and Random Forest (RF) were employed.
• The Ivy algorithm integrated with Bayesian optimization (BOIvy) was used to optimize ML models.
• Performance was evaluated using statistical indices (R², RMSE, U₁, U₂, VAF) and interpretability methods (SHAP, sensitivity analysis).

Main Results:

• The BOIvy-ANN model demonstrated superior predictive performance with optimal statistical indices.
• Water content, curing time, and cement were identified as the most influential factors on CS prediction.
• SHAP and sensitivity analyses provided insights into model interpretability.

Conclusions:

• Optimized machine learning models, particularly BOIvy-ANN, offer a reliable alternative to traditional methods for BPC CS prediction.
• The study highlights the potential of AI techniques in estimating the performance of advanced construction materials.
• Understanding influential factors aids in the efficient design and application of BPC.