Predicting the California Bearing Ratio Applying the Automated Framework of Regression Model

Abstract

The construction of flexible pavement on expansive soil subgrade necessitates the precise determination of the California Bearing Ratio (CBR) value, a crucial aspect of flexible pavement design. However, the conventional laboratory determination of CBR often demands considerable human resources and time. As a result, there is a need to explore alternative methods, such as developing dependable models to estimate the CBR of modified expansive soil subgrade. In this research, a machine learning (ML) model, specifically a Random Forest (RF) machine model, was developed to forecast the CBR of an expansive soil subgrade mixed with sawdust ash, ordinary Portland cement, and quarry dust. The models’ performance was assessed using several error indices, and the findings revealed that the RFAO model exhibited superior predictive capability when compared to the RFDA and RFSM machine models. Specifically, the R2 values for the training and testing data for the RFAO model were 0.9952 and 0.9988, respectively. In addition, RFAO obtained the most suitable RMSE equal to 0.4878. The RFAO model generally indicated an acceptable predictive ability and more desirable generalization ability than the other developed models.

Keywords

• california bearing ratio
• random forest
• dynamic arithmetic optimization algorithm
• slime mould algorithm
• aquila optimizer