Hybrid gradient boosting with meta-heuristic algorithms prediction of unconfined compressive strength of stabilized soil based on initial soil properties, mix design and effective compaction

Highway pavement infrastructure projects which involve soil improvement should be executed, ensuring environmental sustainability. In the present study, guar gum (GG) and lime were assessed for the purpose of soft clay stabilization. The experimental program for the soil stabilization employed a two-stage process. The initial stage involves treatment of the soil with various percentages of lime (3, 5, 7, and 9%) and GG (0.6, 1.0, 1.4, and 1.8%), maintaining the same material acquisition cost and considering curing (0, 7, 14, and 28 days) for the unconfined compressive strength (UCS). In the second experimental stage, a complementary approach in which 3% lime was combined with GG at various percentages (0.1, 0.2, and 0.3%) was employed. The tests conducted include UCS, California bearing ratio (CBR), and strength loss resistance (SLR). Results show that the sole use of lime and GG resulted in significant improvement in the UCS, albeit lime was better. While UCS improved with curing time for the lime-stabilized soil, UCS gain for GG occurred only for up to 7 days curing because biodegradation of GG by microbes in the soil ensues on further curing. Lime-GG stabilization resulted in better UCS and CBR improvement with curing than lime stabilization; however, lime stabilization yielded better SLR. The optimum additive content for strength improvement was obtained at 3% lime + 0.3% GG. Microstructural analysis indicated cementation in the stabilized soil. Predictive models for the UCS were developed based on regression methods. Model evaluation revealed that Gaussian process model provided the best UCS prediction.

The current study applies a soft-computing approach based on the gradient boosting method to predict the unconfined compressive strength (UCS) of sands treated with microbially-induced calcite precipitation (MICP). A 10-fold cross-validation method and hyperparameter tuning are performed to find the optimal architecture of the gradient boosting algorithm. A total of 402 data of unconfined compression tests performed on biocemented sands are utilized in this study. The dataset includes eight input parameters: median sand particle size, uniformity coefficient of sand, initial void ratio, calcium chloride concentration, urea concentration, urease activity, optical density of bacteria, and calcite content. The finding demonstrates that the gradient boosting method outperformed five commonly used machine learning algorithms (artificial neural networks, random forests, k-nearest neighbors, support vector regression, and decision trees) in predicting the UCS of biocemented sands. Using the gradient boosting, the predicted UCS has a strong correlation with the actual values (R2 = 0.95). Moreover, a series of correlation and feature importance analyses are carried out over the dataset. The relationships between unconfined compressive strength, calcite content, and initial void ratio are discussed within the article. Furthermore, some guidelines are provided for assessing the effect of environmental factors on the UCS of biocemented sands. For further study, the limitations of this study regarding the insufficiency of data for correlation and environmental modification are addressed.

The unconfined compressive strength (Qu) is one of the most important criteria of stabilized soil to design in order to evaluate the effective of soft soil improvement. The unconfined compressive strength of stabilized soil is strongly affected by numerous factors such as the soil properties, the binder content, etc. Machine Learning (ML) approach can take into account these factors to predict the unconfined compressive strength (Qu) with high performance and reliability. The aim of this paper is to select a single ML model to design Qu of stabilized soil containing some chemical stabilizer agents such as lime, cement and bitumen. In order to build the single ML model, a database is created based on the literature investigation. The database contains 200 data samples, 12 input variables (Liquid limit, Plastic limit, Plasticity index, Linear shrinkage, Clay content, Sand content, Gravel content, Optimum water content, Density of stabilized soil, Lime content, Cement content, Bitumen content) and the output variable Qu. The performance and reliability of ML model are evaluated by the popular validation technique Monte Carlo simulation with aided of three criteria metrics including coefficient of determination R2, Root Mean Square Error (RMSE) and Mean Square Error (MAE). ML model based on Gradient Boosting algorithm is selected as highest performance and highest reliability ML model for designing Qu of stabilized soil. Explanation of feature effects on the unconfined compressive strength Qu of stabilized soil is carried out by Permutation importance, Partial Dependence Plot (PDP 2D) in two dimensions and SHapley Additive exPlanations (SHAP) local value. The ML model proposed in this investigation is single and useful for professional engineers with using the mapping Maximal dry density-Linear shrinkage created by PDP 2D.

This study investigated the prediction of unconfined compressive strength (UCS), a common measure of soil’s undrained shear strength, using fundamental soil characteristics. While traditional pavement subgrade design often relies on parameters like the resilient modulus and California bearing ratio (CBR), researchers are exploring the potential of incorporating more easily obtainable strength indicators, such as UCS. To evaluate the potential effectiveness of UCS for pavement engineering applications, a dataset of 152 laboratory-tested soil samples was compiled to develop predictive models. For each sample, geotechnical properties including the Atterberg limits, liquid limit (LL), plastic limit (PL), water content (WC), and bulk density (determined using the Harvard miniature compaction apparatus), alongside the UCS, were measured. This dataset served to train various models to estimate the UCS from basic soil parameters. The methods employed included multi-linear regression (MLR), multi-nonlinear regression (MNLR), and several machine learning techniques: backpropagation artificial neural networks (ANNs), gradient boosting (GB), random forest (RF), support vector machine (SVM), and K-nearest neighbor (KNN). The aim was to establish a relationship between the dependent variable (UCS) and the independent basic geotechnical properties and to test the effectiveness of each ML algorithm in predicting UCS. The results indicate that the ANN-based model provided the most accurate predictions for UCS, achieving an R2 of 0.83, a root-mean-squared error (RMSE) of 1.11, and a mean absolute relative error (MARE) of 0.42. The performance ranking of the other models, from best to worst, was RF, GB, SV, KNN, MLR, and MNLR.

Though machine learning (ML) approaches have proliferated in the mechanical properties prediction of cemented paste backfill (CPB), their applications have not reached the peak potential due to the lack of more robust techniques. In the present contribution, the state-of-the-art ensemble learning method was employed for improved estimation of the unconfined compressive strength (UCS) of CPB. 126 UCS tests were conducted on two new tailings to provide an enlarged dataset. Tree-based ML approaches, namely, regression tree (RT), random forest (RF), and gradient boosting regression tree (GBRT), were chosen to be individual ML approaches. The ensemble learning framework was used to combine the optimum individual regressors by means of GBRT. 5-fold cross-validation was used as the validation method and the performance was evaluated using correlation coefficient (R). Hyper-parameters tuning was conducted using particle swarm optimization (PSO). The results show that the best training set size was 70%. PSO was robust in the hyper-parameters tuning since the R value between experimental and predicted UCS on the training set was progressively increased. The ensemble learning can be used to improve the UCS prediction of CPB. The R values between experimental and predicted UCS obtained by RT, RF, GBRT, the ensemble GBRT regressors were 0.9442, 0.9507, 0.9832, and 0.9837, respectively. The method presented in this study extends recent efforts for UCS prediction of CPB and can significantly accelerate the CPB design.

ABSTRACTThe utilization of cemented tailings backfill (CTB) presents distinct advantages in managing tailings and underground mining voids, occasionally incorporating coarse aggregate. In this study, the particle swarm optimization (PSO) algorithm was employed to optimize the extreme gradient boosting (XGBoost) model for predicting the unconfined compressive strength (UCS) of CTB containing coarse aggregate (CTBCA). Additionally, feature importance was compared and analyzed. The findings indicate that the PSO‐XGBoost model exhibits high accuracy on the test set, with a root mean square error (RMSE) of 0.091, a mean square error (MSE) of 0.008, and a coefficient of determination (R2) of 0.999. The predicted values demonstrate a high degree of consistency with the actual results, exhibiting minimal errors that follow a normal distribution. The feature importance analysis reveals that the cement‐sand ratio holds the highest importance score and exerts a significant influence on the UCS prediction. In descending order of impact, the next most significant factors are curing age, slurry concentration, and the coarse aggregate ratio. The proposed PSO‐XGBoost model effectively reduces the UCS measurement cycle while maintaining prediction accuracy. Thus, this model has the potential to provide a fast and efficient method for predicting the UCS of CTBCA.

This paper seeks to develop an interpretable Machine Learning (ML) model for predicting the unconfined compressive strength (UCS) of cohesive soils stabilized with geopolymer at 28 days. Four models including Random Forest (RF), Artificial Neuron Network (ANN), Extreme Gradient Boosting (XGB), and Gradient Boosting (GB) are built. The database consists of 282 samples collected from the literature with three different types of cohesive soil stabilized with three geopolymer categories including Slag-based geopolymer cement, alkali-activated fly ash geopolymer and slag/fly ash-based geopolymer cement. The optimal model is selected by comparing their performances with each other. The values of hyperparameters are tuned by Particle Swarm Optimization (PSO) algorithm and K-Fold Cross Validation. Statistical indicators show the superior performance of the ANN model with three metrics performance such as coefficient of determination R2 = 0.9808, Root Mean Square Error RMSE = 0.8808 MPa and Mean Absolute Error MAE = 0.6344 MPa. In addition, a sensitivity analysis was performed to determine the influence of different input parameters on the UCS of cohesive soils stabilized with geopolymer. The order of feature effect can be ordered in descending order using the Shapley additive explanations (SHAP) value as follows: Ground granulated blast slag content (GGBFS) > Liquid limit (LL) > Alkali/Binder ratio (A/B) > Molarity (M) > Fly ash content (FA) > Na/Al > Si/Al. The ANN model can obtain the best accuracy using these seven inputs. LL has a negative correlation with the growth of unconfined compressive strength, whereas GGBFS has a positive correlation.

Evaluating compressive strength of concrete made with recycled concrete aggregates using machine learning approach

Each type of soil has different optimal soil stabilisation additive content. To design the optimal soil stabilisation component, reliable and efficient models are required. The study proposes the Machine Learning (ML) model Support Vector Regression (SVR) to predict the Unconfined Compressive Strength (UCS) of stabilised soil. To be able to deliver optimal performance, five metaheuristic algorithms: Simulated Annealing (SA), Random Restart Hill Climbing (RRHC), Particle swarm optimisation (PSO), Hunger Games Search (HGS) and Slime Mould Algorithm (SMA) are integrated with the SVR model. To explore the effect of the number of inputs on the model’s performance, the data was divided into two scenarios of input variable number. ML models are evaluated by K-Fold and numerical indicators R 2, RMSE and MAE. The results show that in Scenario 1, the SVR-HGS model has a higher predictive performance than other predictive models. While in Scenario 2, the SVR-PSO model gives better performance than the remaining predictive models. SHapley Additive exPlanation (SHAP) and Partial Dependence Plots 2D (PDP) were used to gain insight into the effects of variables on UCS, and the effects of cement and lime on the variables. Obtaining variables that have an important influence on the variation of stabilised soil UCS, in which cement is considered the most significant variable. The detection of A-line value is a relatively important predictor of UCS. At a suitable A-line value, it is possible to reduce the content of chemical stabilising agents (cement, lime) while maintaining the UCS value at a relative threshold.

Climate signals, driven by complex interactions and nonlinear relationships, shape weather patterns and long-term trends, complicating the identification of dominant drivers due to collinearity. This study investigates the consistency and uncertainty of machine learning (ML) techniques for feature importance in climate science, comparing SHapley Additive exPlanations (SHAP), Partial Dependence Plots (PDPs), and gain-based feature importance from Extreme Gradient Boosting (XGBoost). SHAP’s integration with Feed Forward Neural Networks (FFNN) and XGBoost is evaluated to assess model-specific uncertainties. Using winter precipitation data from Ohio, USA, as a case study, the relative contributions of global warming (GW) and the Interdecadal Pacific Oscillation (IPO) to precipitation changes are quantified. Results show GW consistently ranks higher than IPO in at least 60% of stations across all methods, with SHAP and PDPs agreeing in 89% of stations. Global SHAP importance from FFNN and XGBoost aligns in 82% of stations, with GW contributing 15% more than IPO on average, though disagreements in 18% of stations highlight model-dependent uncertainties. Temporal analysis using SHAP values indicates a moderate discrepancy in feature importance between FFNN and XGBoost models (Pearson correlation ≈ 0.5), despite their consensus on the increasing dominance of GW in recent decades, contributing to wetter winters. Regression analysis further confirms that GW accounts for approximately 70% of the multi-decadal variability in winter precipitation across Ohio, with PDPs indicating a strong monotonicity (ρ = 0.94) between warming levels and precipitation increase. PDPs visualize marginal effects but struggle with interactions, while gain-based methods tend to favor features with a greater number of effective split points that reduce loss. SHAP, though robust for ranking, varies with the base model. An ensemble framework is proposed, demonstrating the value of combining these ML techniques complementarily to account for uncertainties and enhance interpretability. This study highlights the importance of addressing methodological uncertainties in feature importance rankings to provide robust insights for climate modeling.

Background The use of incinerated bottom ash (IBA) as a sustainable construction material offers potential environmental benefits but introduces complex interactions with cement chemistry. Magnesium phosphate cement (MPC), known for its rapid hardening and superior bonding, can be optimized through the controlled incorporation of IBA. However, limited studies have addressed how the chemical components of IBA affect the compressive strength of MPC, particularly using data-driven approaches. Methods A database of 396 experimental samples was compiled from previous studies considering mix proportions, oxide compositions, and curing conditions. Four ensemble machine learning algorithms—Extreme Gradient Boosting (XGB), Light Gradient Boosting (LGB), Gradient Boosting Regressor (GBR), and Random Forest (RFR)—were employed to predict compressive strength. Model robustness was validated through 5-fold cross-validation. Feature interpretation was achieved using SHapley Additive exPlanations (SHAP) and Partial Dependence Plots (PDP) to quantify individual and interactive effects of chemical and physical parameters. Results The XGB model achieved the highest predictive accuracy, with mean training and testing R2 values greater than 0.90 and 0.80, and the lowest mean absolute percentage error of 16.71%. SHAP analysis identified curing age as the most dominant factor, followed by FA/C, W/C, and MgO/PO4 ratios. IBA content and specific oxides such as Fe2O3 and Al2O3 contributed positively to strength within optimal ranges. PDP confirmed nonlinear dependencies, indicating a 26% reduction in strength as W/C increased from 0.1 to 0.6, while extended curing up to 28 days improved performance substantially. Conclusion The integration of SHAP and PDP provided a transparent interpretation of feature interactions in IBA-modified MPC. The developed XGB model demonstrated strong generalization and interpretability. The combined modeling approach offers a reliable predictive framework for optimizing IBA incorporation in sustainable binder systems and advancing eco-efficient material design.

The growing environmental challenges associated with plastic waste disposal and the need for sustainable pavement construction practices have prompted significant research interest in incorporating recycled plastics into asphalt mixtures. However, accurately predicting the performance characteristics of plastic-modified asphalt mixtures, particularly Marshall Stability (MS) and Marshall Flow (MF), remains a critical yet challenging task due to complex nonlinear relationships between mixture constituents. This study addresses this issue by developing reliable predictive models using machine learning techniques including Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), Extreme Gradient Boosting (XGB), and Light Gradient Boosting Machine (LGBM), further optimized through Particle Swarm Optimization (PSO). A comprehensive dataset comprising 210 samples of plastic-modified asphalt mixtures was utilized, incorporating inputs such as plastic content and size, bitumen content, maximum aggregate size, mixing temperature, and compaction effort (number of blows), to predict MS and MF as outputs. Results showed that the PSO-optimized XGB model achieved the highest accuracy, yielding R2 values of 0.82 for MS and 0.83 for MF. Model interpretability was enhanced using advanced techniques such as SHapley Additive exPlanations (SHAP), Partial Dependence Plots (PDP), Individual Conditional Expectation (ICE) plots, and Taylor diagrams, quantitatively highlighting optimal plastic particle sizes (2.5–4 mm), bitumen content (5.3–5.5%) and plastic content (20–30%). These findings provide actionable insights that support safer and longer-lasting pavements, promote the sustainable reuse of waste plastics, and enable cost-effective mix design strategies for modern asphalt construction.

This study investigates the effectiveness of inclined double cutoff walls installed beneath hydraulic structures by employing five machine learning models: Random Forest(RF), Adaptive Boosting(AdaBoost), eXtreme Gradient Boosting(XGBoost), Light Gradient Boosting Machine(LightGBM), and Categorical Boosting (CatBoost). A comprehensive dataset of 630 samples was gathered from previous studies, including key input variables such as the relative distance between the cutoff wall and the structure's apron width (L/B), the inclination angle ratio between downstream and upstream cutoffs (θ2/θ1), the depth ratio of downstream to upstream cutoff walls (d2/d1), and the relative downstream cutoff depth to the permeable layer depth (d2/D). Outputs considered were the relative uplift force (U/Uo), the relative exit hydraulic gradient (iR/iRo), and the relative seepage discharge per unit structure length (q/qo). The dataset was split with a 70:30 ratio for training and testing. Hyperparameter optimization was conducted using Bayesian Optimization (BO) coupled with five-fold cross-validation to enhance model performance. Results showed that the CatBoost model demonstrated superior performance over other models, consistently yielding high R2 values, specifically surpassing 0.95, 0.93, and 0.97 for U/Uo, iR/iRo, and q/qo, respectively, along with low RMSE scores below 0.022, 0.089, and 0.019 for the same variables. A feature importance analysis is conducted using SHapley Additive exPlanations(SHAP) and Partial Dependence Plot (PDP). The analysis revealed that L/B was the most influential predictor for U/Uo and iR/iRo, while d2/D played a crucial role in determining q/qo. Moreover, PDPs illustrated a positive linear relationship between L/B and U/Uo, a V-shaped impact of d2/d1 on iR/iRo and q/qo, and complex nonlinear interactions for θ2/θ1 across all target variables. Furthermore, an interactive Graphical User Interface(GUI) was developed, enabling engineers to efficiently predict output variables and apply model insights in practical scenarios.

Blast-induced rock fragmentation plays a critical role in mining and civil engineering. One of the primary objectives of blasting operations is to achieve the desired rock fragmentation size, which is a key indicator of the quality of the blasting process. Predicting the mean fragmentation size (MFS) is crucial to avoid increased production costs, material loss, and ore dilution. This study integrates three tree-based regression techniques—gradient boosting regression (GBR), histogram-based gradient boosting machine (HGB), and extra trees (ET)—with two optimization algorithms, namely, grey wolf optimization (GWO) and particle swarm optimization (PSO), to predict the MFS. The performance of the resulting models was evaluated using four statistical measures: coefficient of determination (R2), root mean squared error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). The results indicate that the GWO-HGB model outperformed all other models, achieving R2, RMSE, MAE, and MAPE values of 0.9402, 0.0251, 0.0185, and 0.0560, respectively, in the testing phase. Additionally, the Shapley additive explanations (SHAP), local interpretable model-agnostic explanations (LIME), and neural network-based sensitivity analyses were applied to examine how input parameters influence model predictions. The analysis revealed that unconfined compressive strength (UCS) emerged as the most influential parameter affecting MFS prediction in the developed model. This study provides a novel hybrid intelligent model to predict MFS for optimized blasting operations in open-pit mines.

Prediction of unconfined compressive strength and deformation modulus of weak argillaceous rocks based on the standard penetration test