Enhanced Prediction of California Bearing Ratio (CBR) Values in Geotechnical Engineering Using Decision Tree Algorithm and Meta-Heuristic Optimizations

In geotechnical engineering, the CBR test is an essential evaluation tool that can be used in laboratory and field settings. It is essential for figuring out the resistance properties of subgrade soil, whether it is used as the foundation for retaining wall fills, highway embankments, bridge abutments, or earth dams. CBR values provide a valuable metric for assessing the strength of the soil. This paper presents a novel approach to the accurate prediction of CBR values. Using the DT algorithm, the method creates complex and incredibly accurate predictive models. These models include a wide range of intrinsic soil characteristics, including particle distribution, plasticity, linear shrinkage, and the kind and number of stabilizing additives. The dataset of this study consisted of several variables, including LL, PI, PL, MDD, OMC, OPC, SDA, and QD. The DT algorithm improves forecasting accuracy by establishing significant correlations between these soil properties and CBR values. The study incorporates two state-of-the-art meta-heuristic algorithms, the NGO and the EOS, to further improve the predictive model's accuracy. Three unique models are produced by this framework: DTEO, DTNG, and a hybrid DT model. Out of all of them, the DTEO model performs exceptionally well, exhibiting excellent prediction abilities and remarkable generalization. Its performance is rigorously assessed using a range of soil types derived from earlier stabilization tests' outcomes. The DTEO model's remarkable R2 values of 0.996 during the training phase highlight its remarkable accuracy and dependability. Additionally, it achieves an ideal RMSE of 0.732, confirming its accuracy and consistency.