Data-Driven Prediction Model for the Compressive Strength of High-Volume Fly Ash Concrete using Machine Learning Algorithms

Abstract The assessment of concrete properties by conventional methods was uneconomical and time-consuming. The 28-days waiting period to assess its compressive strength may lead to construction delays. The advancement of Machine Learning (ML) technology can help to solve this issue by developing ML model to determine the compressive strength of the concrete, thereby avoiding the waiting period to start the subsequent work in construction. However, there was no prediction model has been developed for assessing the compressive strength of the High-Volume Fly Ash (HVFA) concrete and especially, the model that considers the accelerator as an input variable was not found. This paper introduces a ML prediction model to estimate the compressive strength after 28 days using various algorithms including Linear Regression (LR), k-Nearest Neighbor (kNN), Random Forest (RF), Support Vector Regressor (SVR), Gradient Boost Regressor (GBR), CAT Boost (CATB) regressor and Extreme Gradient Boosting (XGB). It was concluded that CATB model accurately predicts the compressive strength with R 2 of 0.93, RMSE of 4.24 N/mm 2 and MAE of 3.05 N/mm 2 .