California bearing ratio and compaction parameters prediction using advanced hybrid machine learning methods

The modified Proctor compaction parameters and the California bearing ratio (CBR) are critical tests in transportation geotechnics and infrastructure projects. This study employs machine learning (ML) methods to estimate these parameters, which are crucial for guaranteeing the safety and cost-efficiency of infrastructure projects. A dataset of 90 test results gathered from previous studies was employed in the modeling phase. The study applied artificial neural network (ANN), Random forest (RF), Artificial neural network hybridized by neural architecture search (NAS-ANN), and Random forest with neural architecture search (NAS-RF). Based on literature recommendations, six pertinent factors were chosen for the input layer, and the K-fold cross-validation method was utilized for evaluating the effectiveness of all methods. The results indicated that the best-performing model for predicting optimum moisture content (OMC) and maximum dry density (MDD) is NAS-RF, while the optimal model for CBR is NAS-ANN. Additionally, a dependable and user-friendly graphical interface called “ComParaCBR2024” was created in this study. This tool will be highly beneficial for researchers and civil engineers, offering significant time and cost savings in estimating compaction parameters. The findings have demonstrated that the proposed models provide highly effective outputs along with economic benefits.