Evolutionary computing of the California bearing ratio of expansive soil mixed with waste materials and cement

This paper discusses the development of new data-driven models to predict the California bearing ratio (CBR) of expansive soil stabilized using sawdust ash, quarry dust, and cement. The models have been developed using the multi-objective evolutionary polynomial regression analysis utilizing a database of 109 datapoints. The database has been divided into training and testing sets and fed to the evolutionary algorithms. Three models with two, three, and four terms have been developed. These models have been subjected to statistical examination and showed superior accuracy with the obtained mean absolute error (MAE) and root mean square error (RMSE) ranging between 3.4 and 5.7, the mean ranging between 1.01 and 1.08, the 20-index ranging between 0.86 and 0.92, and the obtained coefficient of determination [Formula: see text] ranging between 0.83 and 0.90, respectively. In addition, it has been noticed that the number of terms of the model does not have a noticeable impact on the accuracy, as all of the models scored a similar frequency of cumulative error. Thus, the models could be recommended to be utilized by practitioners, and these models could be examined in the future when new data become available.