Assessment of Artificial Intelligence (AI) Techniques for Estimating the Property of Strength of Geopolymers Concrete

Abstract

Geographic polymer composites (GPCs) are widely studied and favoured. This process involves significant costs and require considerable time investment. Successful research demands innovative methods. This study utilized SVM to predict the compressive strength of GPC, R2, statistics, and k-fold analysis assessed the comparability of all models. Shapley Additive explanations (SHAP) is a model-independent post hoc method that investigates the impact of input factors on GPC CS. Individual ML methods demonstrated lower accuracy in predicting GPC CS compared to ensemble ML approaches. The R2 values for the model and SVM were 0.98 and decreased with the application of ensemble machine learning methods, validating their precision. SHAP identified a more robust positive correlation between GGBS and the compressive strength of GPC. Furthermore, the molarity of NaOH yielded advantageous effects. Both fly ash and 10/20 mm gravel influence the compressive strength of GPC in both positive and negative ways. These components increase CS, while GPC decreases it. Machine learning will assist builders in assessing materials swiftly and cost-effectively.