Machine learning-based prediction of soil compaction parameters
Muharrem Atakan Şentürk, Ertuğrul Ordu, Rabia Korkmaz Tan
Abstract
Abstract In this study, it is aimed to provide significant advantages in terms of time and cost by estimating critical standard compaction parameters such as maximum dry density (MDD) and optimum moisture content (OMC) with machine learning methods instead of traditional laboratory tests. A large dataset including different soil components such as gravel, sand, fine-grained, liquid limit (LL), plastic limit (PL) and plasticity index (PI) was used and algorithms such as decision tree, random forest, gradient boosting and group data processing method (GMDH) were compared. Model performances were evaluated using metrics such as R $$^{2}$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mmultiscripts> <mml:mrow/> <mml:mrow/> <mml:mn>2</mml:mn> </mml:mmultiscripts> </mml:math> (coefficient of determination) and RMSE (root mean square error). The results show that the gradient boosting algorithm achieved high accuracy in estimating optimum moisture content (OMC) with a testing R $$^{2}$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mmultiscripts> <mml:mrow/> <mml:mrow/> <mml:mn>2</mml:mn> </mml:mmultiscripts> </mml:math> value of 0.91, while the random forest algorithm was the most successful model in estimating maximum dry density (MDD) with a testing R $$^{2}$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mmultiscripts> <mml:mrow/> <mml:mrow/> <mml:mn>2</mml:mn> </mml:mmultiscripts> </mml:math> value of 0.92. Machine learning models have been shown to provide faster and lower-cost predictions by reducing the dependency on laboratory tests and offer an effective alternative for soil standard compaction analyses in geotechnical engineering.