Machine Learning Techniques to Predict the Compressive Strength of Metakaolin Blended Concrete

Abstract

Testing the mechanical properties of concrete is costly and time-consuming, often requiring up to 28 days. This study proposes machine learning models to predict the compressive strength of metakaolin- blended concrete, reducing the need for extensive testing. Techniques such as CART, ANN, and GEP were applied using seven predictors: cement, metakaolin, fine aggregate, coarse aggregate, water, superplasticizer, and curing age. Model performance was evaluated with R², RMSE, MSE, and MAPE, and validated using 10-fold cross-validation. Results show that machine learning effectively predicts compressive strength at different curing periods, offering a reliable alternative to traditional testing.