Maximum dry unit weight and optimum moisture content prediction of lateritic soils using regression analysis

Soils compaction with experimental tests is a pivotal facet in the selection of materials for earth constructions. Due to the time limitations and concerns of finishing resources, it is obligate to develop some relationships for predicting compaction parameters such as maximum dry unit weight (γ_dmax) and optimum moisture content (ω_opt) from easily estimated index properties. The purpose is to evaluate the applicability of multivariate adaptive regression splines (MARS) for estimating γ_dmax and ω_opt of lateritic soils. Furthermore, different degrees of interactions of models are employed to have comprehensive, precise, and trustable outputs. The outputs of suggested equations to estimate γ_dmax related to modified proctor compaction test provide proper capability in the modeling procedure. In the training dataset, the value of all criteria for MARS-OI-3 is proper, with the value of 0.9365, 0.4146, and 93.647 for R^2, RMSE, and VAF, respectively. But testing phase’s results are roughly complicated, where scores of MARS-OI-3 equal to 21, bigger than MARS-OI-2 (10) and MARS-OI-4 (17). In summary, MARS-OI-3 outperforms others, where can be known as the suggested equation. The outputs of suggested equations to estimate ω_opt also provide great ability in the modeling. In both phases, the value of all criteria for MARS-OI-2 is proper than MARS-OI-1. Also, scores depict that the score of MARS-OI-2 (15) is about double of MARS-OI-2 (9). So, in spite MARS-OI-1 has justifiable usefulness in the forecasting outline, MARS-OI-2 outperforms it.