Machine Learning in Geotechnics: Predicting Unconfined Compressive Strength of Stabilized Organic Soils Using Hybrid Models
Open Access
- 20 January 2026
- journal article
- Published by Wiley in Advances in Civil Engineering
- Vol. 2026 (1)
Abstract
It is important to predict the unconfined compressive strength (UCS) of stabilized organic soils to be used in designing of foundations. Although machine learning (ML) methods have potential, the literature does not provide overall comparisons. This research suggests a new ML model that uses support vector regression (SVR), multilayer perceptron (MLP), and gradient boosting regression (GBR) and particle swarm optimization (PSO), and hunger games search (HGS) to optimize hyperparameters. The models are trained with 227 measurements of UCS and assessed with the help of cross‐validation. SHapley Additive exPlanations (SHAPs) are used to analyze the importance of the features. The MLP‐HGS and MLP‐PSO model has outstanding performance and the R 2 of 0.999, low root mean square error (RMSE), and mean absolute error (MAE). The most powerful variable that can be identified is cement content, followed by sand, clay, and gravel. The given framework is more accurate and easily interpretable compared to existing studies. The presented innovative method helps to develop ML‐based geotechnical modeling. The results provide information on how the soil stabilization measures can be optimized and the design of stabilized soil foundation made more straightforward.Keywords
This publication has 54 references indexed in Scilit:
- Jacking Force Prediction for Box Jacking Tunnel Considering the Soil Arching EffectInternational Journal for Numerical and Analytical Methods in Geomechanics, 2025
- Three-dimensional strength and deformation characteristics of calcareous sand under various stress pathsBulletin of Engineering Geology and the Environment, 2025
- Research on simpliffied evaluation method for soil-rock mixed slope stability under dam-break flood impactBulletin of Engineering Geology and the Environment, 2025
- Semi‐Analytical Study of Pile–Soil Interaction on a Permeable Pipe Pile Subjected to Rheological Consolidation of Clayey SoilsInternational Journal for Numerical and Analytical Methods in Geomechanics, 2024
- Impact failure and disaster processes associated with rockfalls based on three‐dimensional discontinuous deformation analysisEarth Surface Processes and Landforms, 2024
- Plastic damage prediction of concrete under compression based on deep learningActa Mechanica, 2023
- The prediction of dynamic energy behavior of a Brazilian disk containing nonpersistent joints subjected to drop hammer test utilizing heuristic approachesNeural Computing and Applications, 2022
- Particle swarm optimization algorithm: an overviewSoft Computing, 2017
- Application of Artificial Intelligence to Maximum Dry Density and Unconfined Compressive Strength of Cement Stabilized SoilGeotechnical and Geological Engineering, 2010
- The Nature of Statistical Learning Theory1995