Metaheuristic-optimized machine learning models for predicting compressive strength and assessing sustainability of waste glass powder additive mortars

The research hypothesis of this study is that combining metaheuristic optimization algorithms with ensemble machine learning can significantly improve the accuracy of predicting the compressive strength of mortars containing waste glass powder. The dataset consists of 281 experimental data points compiled from 17 different scientific studies, covering nine input features: cement, sand, glass powder, water, water-to-binder ratio, particle size, curing age, slag, and superplasticizer. The data shows that the PSO-RF model provides the highest predictive performance, achieving an R2 value of 0.943 on test data and 0.841 in real-world experimental validation. Notable findings indicate that curing age and water content are the most critical variables for strength, while a 10% glass powder replacement at 28 days offers the optimal balance between structural performance and environmental sustainability. This information serves as a data-driven decision support system for engineers and researchers to optimize mortar mix designs while reducing carbon emissions and energy consumption.