2026· E3S Web of Conferences· 0 citations· 17 references
TL;DR
This study proposes advanced stacking ensemble machine learning approaches to predict soil Liquidity Index and Undrained Shear Strength and highlights the robust potential of ensemble modeling and tailored optimization in the field of geotechnical engineering.
Abstract
This study proposes advanced stacking ensemble machine learning approaches to predict soil Liquidity Index (
LI
) and Undrained Shear Strength (
Su
). Over 1,550
LI
and 950
Su
measurements from global research were employed. Individual regressors—Multilayer Perceptron (MLP), Random Forest (RF), and AdaBoost—were evaluated alongside ensemble strategies, including Simple Averaging, Weighted Averaging, and stacking using a Support Vector Regression (SVR) meta-learner. Hyperparameter tuning was performed using both Bayesian Optimization (BO) and Particle Swarm Optimization (PSO). AdaBoost achieved the best results for
LI
prediction, while RF yielded the highest accuracy for
Su
. Weighted Averaging ensemble methods produced outstanding predictive performances, achieving
R
2
values of 0.9913 (BO) and 0.9961 (PSO) for
Su
, and 0.9624 (BO) and 0.9338 (PSO) for LI. BO proved superior for
LI
models, while PSO excelled for
Su
models. The results highlight the robust potential of ensemble modeling and tailored optimization in the field of geotechnical engineering.
Accurate prediction of maximum dry density (MDD) and optimum moisture content (OMC) is critical for effective compaction control and earthwork design in geotechnical engineering. Conventional laboratory compaction tests are time-consuming and resource-intensive, motivating the adoption of reliable data-driven prediction models. In this study, a hybrid modeling framework integrating the Rao-1 metaheuristic optimization algorithm with Artificial Neural Network (ANN), Random Forest (RF), and Gradient Boosting (GB) models is proposed for predicting MDD and OMC. A dataset comprising 397 soil samples, characterized by gradation properties and Atterberg limits, is utilized for model development and validation. The Rao-1 algorithm is employed to optimize network weights and model hyperparameters, aiming to enhance convergence behavior and predictive accuracy. Comparative results reveal that Rao-1 optimization consistently enhances model performance across all algorithms and target variables. For MDD prediction, the ANN model achieves an increase in R2 from 0.8512 to 0.9277, accompanied by a reduction in RMSE from 0.4327 to 0.2865. Similarly, the RF and GB models show notable improvements, with optimized R2 values reaching 0.9176 and 0.9213, respectively. For OMC prediction, the Rao-1 optimized ANN exhibits the highest accuracy, improving R2 from 0.8234 to 0.9245 and reducing RMSE from 0.4677 to 0.3071, while optimized RF and GB models also demonstrate substantial error reductions. Furthermore, SHapley Additive exPlanations (SHAP) and the Cosine Amplitude Method (CAM) were integrated to enhance model interpretability, enabling transparent evaluation of feature contributions and providing deeper insight into the influence of geotechnical parameters. Overall, the proposed Rao-1-based hybrid framework significantly enhances predictive accuracy and error minimization compared to conventional models. The results confirm the robustness and effectiveness of Rao-1 optimization in data-driven soil compaction modeling, offering a practical decision-support tool for process innovation and the preliminary estimation of compaction parameters, rather than replacing standardized laboratory testing.
Bayram Ateş, J. Tiang, M. A. Eirgash et al.· Scientific Reports· 0 citations
This study applies several machine learning models, including Gradient Boosting Regressor (GBR), CatBoost (CB), Support Vector Machine (SVM), Random Forest (RF), and AdaBoost (AB), to predict the Marshall stability of Stone Mastic Asphalt (SMA) mixtures incorporating steel slag as a partial replacement for conventional coarse aggregates. The dataset consists of 144 Marshall test samples, with input variables including specific gravity, penetration, flash point temperature, softening point temperature, bitumen content, cellulose fiber content, and steel slag content, while Marshall stability is considered as the output variable. The results indicate that all models achieved high predictive accuracy, with CatBoost providing the best performance, achieving R² = 0.936, RMSE = 0.227, and MAE = 0.323 on the validation dataset. Additional analyses, including residual analysis and SHAP-based interpretation, suggest the robustness and interpretability of the model within the investigated dataset of the model. These findings highlight the strong potential of CatBoost for accurately predicting the mechanical performance of SMA mixtures and support decision-making in mixture design in pavement engineering.
T. Nguyen, Hoang-Long Nguyen, N. Trần et al.· Journal of Science and Trans...· 0 citations
This study presents three advanced machine learning models: the evolutionary Gaussian process inference model, the artificial satellite search algorithm–moment balance machine (ASSA-MBM), and the Operation Rain Forest (ORF), which are designed to predict the maximum reinforcement load in geosynthetic-reinforced soil structures. These models were developed to enhance both predictive accuracy and model interpretability by incorporating state-of-the-art optimization algorithms and explainable machine learning frameworks. A comprehensive evaluation was conducted using 10-fold cross-validation, and the proposed models were benchmarked against previously developed AI models from literature, as well as traditional and semiempirical approaches such as Rankine, Coulomb, and
K
-stiffness. Among the proposed models, ASSA-MBM consistently achieved the best performance, recording the lowest testing root mean squared error (0.617), the highest correlation coefficient (
R
=
0.918
), and the highest reference index (
RI
=
0.951
). Additionally, the ORF model offers transparency by generating mathematical regression equations, which are crucial in geotechnical engineering.
Min-Yuan Cheng, Akhmad F. K. Khitam, Jia-Wang Liou· Journal of computing in civi...· 0 citations
This study utilizes robust machine learning (ML) techniques to predict compressive strength of fly-ash-based geopolymer concrete (GPC), a sustainable replacement for traditional concrete. Popular ensemble models, specifically Random Forest (RF) and Extreme Gradient Boosting (XGB), were taken as base models and were hybridized using metaheuristic algorithms (Particle Swarm Optimization (PSO) and Grey Wolf Optimizer (GWO) for hyperparameter optimization. A 5 × 5 nested cross-validation (nCV) approach was adopted, where inner folds were used for hyperparameter tuning, and outer folds for unbiased performance evaluation for the limited 273-sample dataset. The findings revealed that GWO-XGB outperformed other hybridized processes, with aggregated R2 and RMSE of 0.9661 ± 0.011 and 3.0401 ± 0.5363, respectively, in the testing phases. The performance ranking for both the training and testing phases was: GWO-XGB > PSO-XGB > GWO-RF > PSO-RF. Further, SHAP analysis was performed on models obtained from the best-tuning process, which identified Curing Period (CTP) and Curing Temperature (CTR) as the most critical parameters influencing FA-GPC strength. The best-optimized model was also used to build a graphical user interface (GUI). This work offers a reliable framework, not only for predicting CS but also for demonstrating how feature interactions contribute to strength development, providing insights into effective ML use in concrete technology.
Subodh Subedi, Ajaya Subedi, S. Adhikari et al.· Bulletin of Computational In...· 0 citations
The current study combines numerical modelling and machine learning to identify the stability of applications in nail reinforced slope study. PLAXIS LE was used to develop different slopes having different soil properties including various values for cohesion (5, 10, 15 kPa), angle of internal friction (20°, 25°, 30°), unit weight (17, 18, 19 N/m³), and slope angle (30°, 35°, 40°, 45°, 50°, 60°, 70°). Safety Factors (FOS) prediction models such as Random Forest (RF), Linear Regression (LR), and K-Nearest Neighbors (KNN) have been developed using the parameters included in the study. The Random Forest model has shown a superior performance among the other models with the lowest Mean Absolute Error (MAE: 0.053) and Mean Squared Error (MSE: 0.006), taking into consideration the highest value of R² (0.957) and Adjusted R² (0.951) to indicate a better predictive accuracy. With R² values of 0.903 and 0.920, respectively, Linear Regression and KNN also showed considerable strength of results. The results mentioned above show the bright future of machine learning models with Random Forest in predicting slope stability and contribute to refining nail reinforcement strategies. It shall also provide an input for developing cost-effective and robust slope rehabilitation measures in a geotechnically unfriendly environment.
This paper presents an interpretable machine-learning framework for predicting the splitting strength (ST) of asphalt concrete and supporting data-driven mixture design. A database consisting of 296 samples was established, and 14 input variables related to asphalt properties, aggregate gradation, and fiber characteristics were selected for modeling. Six machine-learning models, namely TabPFN, ANN, SVR, RF, XGBoost, and LightGBM, were developed and compared. Hyperparameter optimization was performed for five models using NSGA-II, while TabPFN was directly applied with its default configuration. The results show that all six models achieved satisfactory predictive capability, whereas TabPFN delivered the best overall performance on the testing set, with the lowest RMSE of 0.28, MAE of 0.21, MAPE of 18.01%, MAD of 0.14, the highest R^2 of 0.88, and the highest composite score of 0.91. SHAP analysis further revealed that nine dominant variables accounted for 92.0% of the total average contribution, among which Ag9.5, FT, Ag4.75, AC, and Du were the most influential. In addition, favorable parameter ranges for improving ST were quantified, such as Ag9.5<66.8%, Ag4.75<45.0%, AC<5.4 wt.%, AV<3.6%, and Du>134.7 cm. Finally, a GUI platform integrating prediction and SHAP-based explanation was developed to improve the accessibility and practical applicability of the proposed framework.
J. Xing, Xiao Tan, Dongzhan Jin et al.· 0 citations