Predicting Compressive Strength of Color Pigment Incorporated Roller Compacted Concrete via Machine Learning Algorithms: A Comparative Study

Abstract

Due to its low albedo, traditional asphalt pavement contributes to the urban heat island effect. Color pigment added roller compacted high performance concrete is a novel approach to reducing the urban heat island effect through the use of paving materials. In this study, color pigment added roller compacted concrete specimens were produced and evaluate via the machine learning algorithms. Predicting compressive strength of concrete by utilization of machine learning methods is highly preferred method by scholars and professionals since ingredients' resources are intensive and time consuming. This research focused to predict the compressive strength of color pigment incorporated roller compacted concrete by applying multiple linear regression (ML), gradient boosting (GB), random forest (RF), support vector machines (SVM), artificial neural network (ANN) and bagging algorithms (BGG). A comprehensive database containing coarse aggregates, fine aggregate, water, cement and pigment amounts and density, age information as input parameters. The analysis results reveal that Bagging algorithm was able to obtain more satisfactory results than the other algorithms in predicting compressive strength (CS) of color pigment incorporated roller compacted concrete. In this algorithm, root mean square error (RMSE) was determined to be 1.53, R-2 to be 0.962, mean absolute error (MAE) to be 0.916, and mean absolute percentage error (MAPE) to be 0.033. ANN algorithm showed significant accuracy in prediction process with RMSE of 1.725, R-2 of 0.949, MAE of 1.144, and MAPE of 0.040. The lowest accuracy was obtained in SVM algorithm with RMSE of 26.910 R-2 of 0.512, MAE of 3.981, and MAPE of 0.040. Therefore, the present study can provide an efficient option for estimating the of color added Roller compacted concrete for pavements.