An Artificial Intelligence-Based Model for Geopolymer Concrete Strength Prediction

Abstract

Geopolymer concrete (GPC) has emerged as a sustainable alternative to conventional concrete, offering reduced carbon emissions and enhanced mechanical properties. However, variability in compressive strength due to material composition poses challenges to its broader adoption. Traditional evaluation methods are often time-consuming and resource-intensive, necessitating the development of precise and efficient predictive tools. This study introduces the optimized least squares moment balanced machine with feature selection (OLSMBM-FS), an advanced AI-based model for accurately predicting GPC compressive strength. The model incorporates backpropagation neural networks (BPNN) for weight assignment, least squares support vector machines (LSSVM) for hyperplane optimization, and the optical microscope algorithm (OMA) for hyperparameter tuning. The study employs a systematic dataset, implementing normalization and feature selection techniques to improve the accuracy and efficiency of the model training process. The OLSMBM-FS was validated using 10-fold cross-validation and demonstrated superior performance compared to other machine learning models. It achieved the lowest RMSE (4.279), MAE (2.291), and MAPE (6.59%), alongside the highest R (0.901) and R² (0.813), confirming its robustness and predictive accuracy. These findings highlight the potential of OLSMBM-FS as a reliable tool for predicting GPC compressive strength, supporting its broader application in sustainable construction practices.