Evaluation of Standard Compaction Parameters of Lateritic Soils Using Regression Analysis

Experimentally, soils compaction is a significant facet in choosing materials for the construction of ground. Due to scheduling constraints and concern for finishing sources, it is necessary to establish relationships to predict compaction parameters such as optimum moisture content (\({\omega }_{\mathrm{opt}}\)) and highest dry unit weight (\({\gamma }_{\mathrm{dmax}}\)) derived from simple index attributes. This research aims to produce practical regression models using multivariate adaptive regression splines (\(MARS\)) and support vector regression (\(SVR\)) for predicting \({\gamma }_{\mathrm{dmax}}\) and \({\omega }_{\mathrm{opt}}\) of standard lateritic soils. In the \(SVR\) analysis, the whale optimization algorithm (\(WOA\)) was applied to the \(SVR\) to determine the key parameters of \(SVR\). The result of the proposed equations to predict \({\gamma }_{\mathrm{dmax}}\) relevant to the standard lateritic soils’ supervisor compression test depicts appropriate potential in the modeling process. In the testing and training stage, all criteria’s value for MARS-OI-3 is better than MARS-OI-2, with value of 0.9146, and 91.46 for \({R}^{2}\) and ‘variance accounted factor (\(VAF\)), respectively. Summated scores present that the score of MARS-OI-3 is 13, larger than MARS-OI-2. The result of developed equations for predicting \({\omega }_{\mathrm{opt}}\) related to the standard proctor compaction test of lateritic soils present suitable capability during the modeling process. In both the training and testing phase, the value of all criteria for MARS-OI-2 is better than MARS-OI-1. All in all, to predict \({\gamma }_{\mathrm{dmax}}\), the MARS-OI-3 model, and \({\omega }_{\mathrm{opt}}\), the MARS-OI-2 equation can be recognized as the proposed equation. The developed WOA-SVR model carried out a great performance in the prediction procedure of \({\gamma }_{\mathrm{dmax}}\) and \({\omega }_{\mathrm{opt}}\), by gaining the best values of all indices with respect to MARS models.