Extreme Gradient Boosting Research Articles

An interpretable surrogate model for H2S solubility forecasting in ionic liquids based on machine learning

Here we investigated four different ML-based models, i.e., gaussian process regression (GPR), extreme gradient boosting (i.e., XGBoost), random forest (RF), and support vector machine (SVM), for predicting the solubility of H2S in various ionic liquids (ILs). The dataset was divided into training and testing sets in an 80:20 ratio while the model performance for all models were evaluated using the coefficient of determination (R2), mean absolute error (MAE), and root mean square error (RMSE). Overall, all models effectively predicted H2S solubility, albeit with varying degrees of performance. The GPR provides the best performance, with R2 of 0.9918, MAE of 0.0090, and RMSE of 0.0147. Following this is the XGBoost model with an R2 value of 0.9827, MAE of 0.0155, and RMSE of 0.0213. The RF model displayed slightly lower performance, with an R2 value of 0.9395, MAE of 0.0261, and RMSE of 0.0398 while the lowest performance was demonstrated by the SVM model, which gave an R2 value of 0.9036, MAE of 0.0402, and RMSE of 0.0508. We used SHAP analysis and identified the pressure, temperature, Estate_VSA3, Estate_VSA5, and MinEStateIndex as the top five dominant input features in our model interpretation. In a nutshell, this work presents new insights into the molecular characteristics that affect the solubility of H2S in ILs, paving future research path in this field.

Predicting the Recurrence of Ovarian Cancer Based on Machine Learning.

Recurrence is the main factor for poor prognosis in ovarian cancer, but few prognostic biomarkers were reported. In this study, we used machine learning methods based on multiple biomarkers to develop a specific prediction model for the recurrence of ovarian cancer. A total of 277 ovarian cancer patients were enrolled in this study and randomly classified into training and testing cohorts. The prediction information was obtained through 47 clinical parameters using six supervised clustering machine learning algorithms, including K-Nearest Neighbor (K-NN), Decision Tree (DT), Random Forest (RF), Adaptive Boosting (AdaBoost), Gradient Boosting Machine (GBM), and Extreme Gradient Boosting (XGBoost). In predicting the recurrence of ovarian cancer, machine learning algorithm was superior to conventional logistic regression analysis. In this study, XGBoost showed the best performance in predicting the recurrence of ovarian cancer, with an accuracy of 0.95. In addition, neoadjuvant chemotherapy, Monocyte ratio (MONO%), Hematocrit (HCT), Prealbumin (PAB), Aspartate aminotransferase (AST), and carbohydrate antigen 125 (CA125) are the most important biomarkers to predict the recurrence of ovarian cancer. The machine learning techniques can achieve a more accurate assessment of the recurrence of ovarian cancer, which can help clinicians make decisions, and develop personalized treatment strategies.

Community identification and carbon storage monitoring of Heritiera littoralis with UAV hyperspectral imaging

The Heritiera littoralis, an important rare semi-mangrove species found in coastal protective forests, plays a crucial role in carbon storage and cycling within mangrove wetland ecosystems. Despite its importance, previous studies have overlooked remote monitoring of its carbon storage. Taking the world’s oldest and largest natural community of H. littoralis in Shenzhen, China, as the study area, this research pioneers the use of UAV hyperspectral imaging to identify the H. littoralis community and estimate its above-ground, below-ground, and total carbon storage. The impacts of five geo-environmental factors (elevation, slope, aspect, vegetation communities, and inland distance from the coastline) on the spatial variability of carbon storage using SHapley Additive exPlanations (SHAP) and multiscale geographically weighted regression (MGWR) methods were also investigated. The results demonstrate that the first derivative bands within the red edge and near-infrared regions, the anthocyanin reflectance index 2 (ARI2) and newly-developed three-band VIs, were sensitive features for community identification and carbon storage estimation within the H. littoralis community. Among the four machine learning models (eXtreme Gradient Boosting (XGBoost), Random Forest (RF), Support Vector Machine (SVM), and Logistic Regression (LR)), the best identification accuracy for the H. littoralis community was achieved by XGBoost. Among the four machine learning models (XGBoost, RF, SVM and kernel ridge regression (KRR)), RF model achieved the best performance in estimating above-ground (R2 = 0.749, RMSE=1.723 kg/m2, EV (explained variance) = 0.704), below-ground (R2 = 0.636, RMSE=0.6 kg/m2, EV=0.606) and total carbon storage (R2 = 0.613, RMSE=2.592 kg/m2, EV=0.597). SHAP and MGWR analysis showed that elevation and inland distance from the coastline were key factors influencing the spatial variability of carbon storage. In conclusion, this study showcases significant advantages in community identification and carbon storage monitoring of rare mangrove communities, providing valuable insights for biodiversity conservation efforts and management within the H. littoralis ecosystem.

A robust approach for bond strength prediction of mortar using machine learning with SHAP interpretability

Application of machine learning (ML) in predicting mortars bond strength contributes to low experimental cost and high accuracy. This study explores the performance of four ML models—Back Propagation Neural Network (BPNN), Support Vector Regression (SVR), Random Forest (RF), and eXtreme Gradient Boosting (XGBoost)—using 304 experimental data points for training and 76 for testing. Among these, the XGBoost model demonstrated the highest predictive accuracy, with R², MAE, and RMSE values of 0.95, 0.06, and 0.13 in the training set, and 0.94, 0.08, and 0.16 in the testing set, respectively. Compared with traditional models, ML approaches, particularly XGBoost, were more effective in providing reliable predictions of bond strength. SHapley Additive exPlanation (SHAP) show that substrate type, ethylene-vinyl acetate (EVA) content and water to binder ratio W/C have the most remarkable effect on bond strength, while the type of cement and admixture shows the least effect. These findings highlight the potential of ML models in optimizing mortar formulations and improving prediction reliability.

Study on the Genesis and Development Trend of Tropical Depressions under Different Large-Scale Backgrounds

Based on self-organizing maps (SOM), large-scale backgrounds associated with tropical depression (TD) genesis over the western North Pacific (WNP) in 1949–2021 are classified into four circulation patterns, monsoon gyre (MG) pattern, monsoon confluence (MC) pattern, monsoon trough (MT) pattern and easterly wave (EW) pattern. TDs generated in the MC pattern has the southernmost average genesis location and the highest development probability, while TDs occurred in the EW pattern are averagely located northernmost and their probability of development is the lowest. TDs formed in the MG, MT and EW patterns are most active in August, whereas in the MC pattern, TD genesis number peaks in October. Advantageous conditions for TD genesis vary in different circulation patterns. The vigorous vorticity "embryo" provides stronger initial disturbances for MG pattern; The strong upper-level divergence and the weak deep-layer VWS provide sufficient dynamic conditions for the MC pattern; The MT pattern possess the highest SST, which supplies an ample supply of heat and moisture; The EW pattern has less beneficial conditions compared with other three patterns. The Extreme Gradient Boosting (XGBoost) method is applied to quantify the relative importance of individual factors to TD development trend. 500-hPa vorticity, 200-hPa divergence and SST are major dynamic and thermal affecting factors for TD development, the importance of which all ranked at top four in the four patterns; VWS plays an indispensable role in TD development for the MC and EW patterns; Comparely,850-hPa vorticity and vertically integrated water vapor flux are not as important as above environmental factors in deciding whether a TD develops.

Satellite retrieval of oceanic particulate organic carbon: Towards an accurate and seamless dataset for the global ocean

Particulate organic carbon (POC) plays crucial roles in the global ocean carbon cycle and the oceanic biological pump. Satellite remote sensing has been demonstrated to be an effective technique for the retrieval of surface oceanic POC concentration. However, the complex spatiotemporal variations of the relationships between POC and oceanic optical properties across different waters posed challenges for accurate retrieval of POC concentration from satellite observations. Additionally, interference factors, such as cloud cover and sun glint, resulted in severe data missing problems and impeding daily coverage of the global ocean. With an attempt to generate accurate, seamless and readily available POC products for the global ocean, this study aimed to develop accurate satellite POC retrieval models for the Moderate Resolution Imaging Spectroradiometer (MODIS) data from both Terra and Aqua satellites, and to explore the possibility of using the empirical orthogonal function interpolation technique (DINEOF) to reconstruct satellite-retrieved POC data to generate gap-free global oceanic POC products. Results showed that the eXtreme Gradient Boosting (XGBoost) method could accurately retrieve POC with R2 approximately 0.80 and RMSE about 0.20 in log10 scale, obviously outperforming the operational blue-to-green band ratio algorithm and the hybrid polynomial algorithm based on two multi-band indices; and the DINEOF method, which could reconstruct approximately 88 % missing pixels for the global ocean, contributed to better revealing the global oceanic POC variations at a daily scale than the satellite-retrieved POC products. Based on the developed models, a suit of long time-series accurate and seamless POC products of the global surface ocean were generated, which is readily available for other applications and should be helpful to investigate the spatiotemporal variations of POC concentrations over global ocean and its roles in the global carbon cycle. The generated seamless products are openly accessible via the DOIs listed in the data availability section.

Long-term AI prediction of ammonium levels in rivers using transformer and ensemble models

This study provides a cutting-edge machine learning approach to forecast ammonium (NH4+) levels in River Lee London. Ammonium concentrations were predicted over several time intervals using a complete dataset that includes temperature, turbidity, chlorophyll, dissolved oxygen, conductivity, and pH. Our technique captures the intricate connections between environmental conditions and ammonium concentrations using developed algorithms, including Temporal Fusion Transformer (TFT), Random Forest (RF) and Extreme Gradient Boosting (XGBoost) levels versus the important factors, considerably improving prediction accuracy. The novel aspect of this study is the utilisation of the TFT model for multi-horizon forecasting, which offers high accuracy and interpretability in hydrological predictions by combining convolutional components with an attention mechanism. The study also demonstrates the effectiveness of the TFT model in capturing short-term fluctuations while retaining accuracy over long time periods, which is a major difficulty in environmental modelling. The models used, have exceptional forecasting skills, predicting 150, 200, 365, 730, and 1095 days based on daily average and 12, 24 and 30 months based on monthly average. This dual-scale model combines flexibility and resilience, making it an effective tool for forecasting both short- and long-term environmental changes. The RF model excelled in long-term forecasts, sustaining high R-squared (R²) (0.97) values and low root mean square error (RMSE) (0.18), and the second best one was the XGBoost with optimiser with R2 of (0.92) and RMSE of (0.25) with forecasting 1095 days. The results also found that whilst the TFT captured the fluctuations in the short-term, it struggled with the longer-term predictions due to data granularity. The XGBoost model did remarkably well in monthly forecasts up to 12 months, maintaining low RSME. The findings also highlight the necessity of proactive water management techniques to reduce the risk of potential ecological effects, including hypoxia and oxygen depletion. The findings support resource managers in addressing prospective ammonium toxicity concerns such as oxygen depletion and ecological stress.

Assessing and explaining rising global carbon sink capacity in karst ecosystems

Global karst vegetation is crucial for carbon sequestration and biodiversity conservation. However, research on the carbon sink capacity of vegetation and its potential driving mechanisms across different geographic types of global karst remains scant. This research delineates the trend characteristics and differences of net primary productivity (NPP), heterotrophic respiration (Rh), and net ecosystem productivity (NEP) across various bedrock and climatic zones within global karst landscapes. Extreme Gradient Boosting (XGBoost), Shapley Additive Explanations (SHAP), Neural Network fitting, and Partial Least Squares Structural Equation Modeling were utilized to investigate the nonlinear impacts and underlying mechanisms exerted by the explanatory variables on NPP, Rh, and NEP. The findings indicate that from 1981 to 2019, the rates of increase in NPP (2.02 gC·m⁻2 yr⁻1), Rh (1.47 gC·m⁻2 yr⁻1), and NEP (0.54 gC·m⁻2 yr⁻1) were marginally higher in global karst regions compared to non-karst areas. Additionally, when contrasted with various bedrock types, the differences in mean values and trends of these three indicators were more pronounced across different climatic zones of karst areas. XGBoost effectively captures the nonlinear relationship between NEP and its explanatory variables, while SHAP provides a detailed explanation of the contribution and interaction of these variables. The leaf area index (LAI) was the major contributor to NEP. Additionally, the nonlinear effects of individual critical drivers on NEP varied across different aridity gradients. The cascade effect of explanatory variables on NEP follows a pathway where geography affects climate, climate change subsequently alters vegetation characteristics, and these changes in vegetation directly drive NPP and Rh, ultimately leading to an indirect impact on NEP. This impact pathway exhibits minimal difference across different bedrock and climate types. This study enhances our comprehension of the role of vegetation carbon sinks and their underlying mechanisms across various bedrock and climate types in global karst regions.

Quantitative risk assessment for overtopping of earth-fill dams in Japan using machine learning algorithms

Earth-fill dams serve as crucial agricultural structures in Japan and act as buffers against flooding. However, their failure often tends to cause even greater downstream damage. Consequently, there is an urgent need for a quantitative assessment of the risks to earth-fill dams posed by disasters. The current detailed method of assessment is complicated, labour-intensive, and costly; hence, constructing risk surrogate models will greatly reduce the workload. This study employs two machine learning methods, GPR (Gaussian Process Regression) and XGBoost (eXtreme Gradient Boost), to develop surrogate models for assessing the damage cost and overtopping probability for 70 earth-fill dams in Okayama and Hiroshima prefectures, Japan. The predictive performance of each model was quantified by comparing the results against those of the detailed method. From the results, XGBoost demonstrates superior performance compared to GPR based on the comparison of coefficient of determination (R2) and root mean square error (RMSE). To clarify the extent to which the variables influence the XGBoost model, the SHapley Additive exPlanations (SHAP) algorithm was implemented. It offers an efficient and interpretable avenue for earth-fill dam risk assessments.

A highly accurate and robust prediction framework for drilling rate of penetration based on machine learning ensemble algorithm

The rate of penetration (ROP) is a key indicator of drilling efficiency. Many researchers have explored the application of machine learning in ROP prediction. However, few studies have focused on the robustness of the constructed models, and developing a ROP prediction model that can achieve both high accuracy and strong robustness remains a challenge. This paper introduces a novel machine learning approach to constructing a ROP prediction model through ensemble learning algorithms. The model is based on field data from oilfields, incorporating 16 input parameters that influence ROP. First, the feasibility of the collected dataset is verified using correlation analysis. Then, ROP prediction models are developed based on various machine learning algorithms, including Decision Tree Regression (DTR), Random Forest (RF), eXtreme Gradient Boosting (XGB), Light Gradient Boosting Machine (LGBM), Support Vector Regression (SVR), and Backpropagation Neural Network (BPNN). By comparing the performance of these models under noise levels of 0%, 1.7%, and 5.1%, RF, LGBM, XGB, and SVR are selected as base learners. These base learners are then combined to construct multiple ensemble models, and the performance of the optimal ensemble model is evaluated under varying noise levels. The results show that the prediction error of the optimal model remains within 10%, and R2 is greater than 0.96. Finally, the Shapley Additive Explanations (SHAP) method is applied to perform interpretability analysis on the optimal ROP prediction model, examining the impact of different input factors on the model's predictive performance. Compared to single models and other ensemble models, the proposed ensemble model not only achieves higher accuracy but also demonstrates strong robustness and generalization capability.

Two-step machine learning-aided two-stage hydrothermal liquefaction of biomass for bio-oil upgrading to lower nitrogen content: Experimental verification and parameter optimization

The two-stage hydrothermal liquefaction (HTL) is a promising process for converting biomass into low-nitrogen bio-oil. However, for two-stage HTL, there is a considerable lack of application of artificial intelligence tool. In this study, a two-step machine learning (ML) model was developed to optimizing bio-oil production of two-stage HTL. A total of 644 data points was collected, and a novel feature NH2-C (the mass ratio of amino to carbon) was used to enhancing ML models using Random Forest (RF), Gradient Boosting Regression (GBR), and Extreme Gradient Boosting (XGB) algorithms. The XGB multi-task model demonstrated the best performance (train R2 of 0.98, test R2 of 0.94 and 0.83). The nitrogen content in feedstock is the decisive factor for total nitrogen and nitrogen recovery rate of aqueous phase in HTL stage Ⅰ, while NH2-C for yield and nitrogen content of bio-oil in stage Ⅱ. The two-step ML model performed excellently in 22 validation experiments (with all errors ≤13.68%), and guided the bio-oil preparation with lower nitrogen content (5.46%) from Chlorella. This study provides a new approach for the application of multi-step ML in biomass HTL.

Machine learning-based prediction of adsorption capacity of metal-doped and undoped activated carbon: Assessing the role of metal doping

This research developed five ensemble-based machine learning (ML) models to predict the adsorption capacity of both pristine and metal-doped activated carbon (AC) and identified key influencing features. Results indicated that Extreme Gradient Boosting (XGB) model provided the most accurate predictions for both types of AC, with metal-doped AC exhibiting 1.7 times higher adsorption capacity than pristine AC showing 254.66 and 148.28 mg/g, respectively. Feature analysis using SHAP values revealed that adsorbent characteristics accounted for 53.5 % of the adsorption capacity in pristine AC, while experimental conditions were crucial for metal-doped AC (61.3%), with surface area and initial concentration being the most significant features, showing mean SHAP values of 0.317 and 0.117, respectively. Statistical comparisons of adsorbent characteristics between pristine and metal-doped AC showed that metal doping significantly altered surface area (p-value = 0.0014), pore volume (p-value = 0.0029), and elemental composition (C% (p-value = 3.9513∗10^−7) and O% (p-value = 0.0007)) of AC. Despite the reduction in surface area and consistent pore volume after metal doping, the enhanced adsorption capacity of metal-doped AC was attributed to increased oxygen content from 10.89% to 17.28 % as mean values. This suggests that oxygen-containing functional groups play a critical role in the improved adsorption capacity of metal-doped AC. This research lays the groundwork for optimizing AC adsorbents by identifying key factors in metal-doped AC and suggest further studies on the interaction between specific metal dopants and resulting functional groups to improve adsorption capacity and reduce repeated labor work.

Optimization of drug solubility inside the supercritical CO2 system via numerical simulation based on artificial intelligence approach

In this research paper, we explored the predictive capabilities of three different models of Polynomial Regression (PR), Extreme Gradient Boosting (XGB), and LASSO to estimate the density of supercritical carbon dioxide (SC-CO2) and the solubility of niflumic acid as functions of the input variables of temperature and pressure. The optimization of hyperparameters for these models is achieved using the innovative Barnacles Mating Optimizer (BMO) algorithm. For SC-CO2 density estimation, PR exhibits remarkable accuracy, showing an R-squared value of 0.99207 for data fitting. XGB performs admirably with an R2 of 0.92673, while LASSO model demonstrates good predictive ability, showing an R2 of 0.81917. Furthermore, we assess the models’ performance in predicting the solubility of niflumic acid. PR exhibits excellent predictive capabilities with an R2 of 0.96949. XGB also delivers strong performance, yielding an R-squared score of 0.92961. LASSO performs well, achieving an R-squared score of 0.82094. The results indicated promising performance of machine learning models and optimizer in estimating drug solubility in supercritical CO2 as the solvent applicable for pharmaceutical industry.

Development and external validation of a machine-learning based model to predict pre-sarcopenia in MASLD population: Results from NHANES 2017–2018

Introduction and ObjectivesWith rising prevalence of pre-sarcopenia in metabolic dysfunction-associated steatotic liver disease (MASLD), this study aimed to develop and validate machine learning-based model to identify pre-sarcopenia in MASLD population. Materials and MethodsA total of 571 MASLD subjects were screened from the National Health and Nutrition Examination Survey 2017–2018. This cohort was randomly divided into training set and internal testing set with a ratio of 7:3. Sixty-six MASLD subjects were collected from our institution as external validation set. Four binary classifiers, including Random Forest (RF), support vector machine, and extreme gradient boosting and logistic regression, were fitted to identify pre-sarcopenia. The best-performing model was further validated in external validation set. Model performance was assessed in terms of discrimination and calibration. Shapley Additive explanations were used for model interpretability. ResultsThe pre-sarcopenia rate was 17.51 % and 15.16 % in NHANES cohort and external validation set, respectively. RF outperformed other models with area under receiver operating characteristic curve (AUROC) of 0.819(95 %CI: 0.749, 0.889). When six top-ranking features were retained as per variable importance, including weight-adjusted waist, sex, race, creatinine, education and alkaline phosphatase, a final RF model reached an AUROC being 0.824(0.737, 0.910) and 0.732(95 %CI: 0.529, 0.936) in internal and external validation sets, respectively. The model robustness was proved in sensitivity analysis. The calibration curve and decision curve analysis confirmed a good calibration capacity and good clinical usage. ConclusionsThis study proposed a user-friendly model using explainable machine learning algorithm to predict pre-sarcopenia in MASLD population. A web-based tool was provided to screening pre-sarcopenia in community and hospitalization settings.

Interpreting machine learning models based on SHAP values in predicting suspended sediment concentration

Machine learning (ML) has become a powerful tool for predicting suspended sediment concentration (SSC). Nonetheless, the ability to interpret the physical process is considered the main issue in applying most of ML approaches. In this regard, the current study presents a novel framework involving four standalone ML models (extra trees (ET), random forest (RF), categorical boosting (CatBoost), and extreme gradient boosting (XGBoost)) and their combination with genetic programming (GP). Three metrics (coefficient of correlation (r), root mean square error (RMSE), and Nash–Sutcliffe model-fit efficiency (NSE)) and a more advanced interpretation system SHapley Additive exPlanations (SHAP) are used to assess the performance of these models applied to hydro-climatic datasets for prediction of SSC. The calibration process was based on data from 2016 to 2020, and the validation was done for 2021 data. Further description and application of the framework are provided based on a case study of the Bouregreg watershed. The results revealed that all implemented models are efficient in SSC prediction with NSE, RMSE, and r varying from 0.53 to 0.86, 1.20 to 2.55 g/L, and 0.83 to 0.91 g/L respectively. Box plot diagrams confirm the enhanced performance of these combined models, and the best-performing ones for the four hydrological stations being the combined RF+GP model at the Aguibat Ziar station, the combined XGBoost+GP model at the Ain Loudah station, the CatBoost model at the Ras Fathia station, and the RF model at the Sidi Med Cherif station. The interpretability results showed that flow (Q) and seasonality (S) are the features most impacting SSC. These outcomes indicate that the applied models can extract accurate and detailed information from the interactions between the hydroclimatic factors and the generation of sediment by erosion (output). ML approaches illustrated the good reliability and transparency of the models developed for predicting SSC in a semi-arid setting, offered new perspectives for reducing ML models' "black box" character, and provided a useful source of information for assessing the consequences of SSC on water quality. The SHAP system and exploring other interpretable techniques are recommended to provide further information in future research. In addition, incorporating additional input data could enhance SSC predictions and deepen understanding of sediment transport dynamics.

Machine learning assisted prediction of specific surface area and nitrogen content of biochar based on biomass type and pyrolysis conditions

Predicting and optimizing the physicochemical properties of biochar is crucial for its applications. The characteristics of biomass and pyrolysis conditions are the main factors influencing these properties. However, the numerous components of biomass and the pyrolysis conditions contribute to the substantial challenge in predicting the physicochemical properties, particularly the specific surface area and the nitrogen content of biochar. In this work, machine learning methods including random forest (RF), gradient boosting decision tree (GBDT) and extreme gradient boosting (XGB) (all with R2 exceeding 0.97) were used to predict and analyze specific surface area of biochar (SSA), N content of biochar (N-char), and yield of biochar (Yield-char). Compositions of biomass and pyrolysis conditions were selected as input variables. The partial dependence plot analysis showed the impact way of each influential factor on the target variable and the interactions among these factors in the pyrolysis process. The feature importance of these models indicated that the influencing factors toward predicting three targets (sorted by importance) were specified as follows: pyrolysis temperature, nitrogen content, and fixed carbon for Yield-char; N and ash for N-char; ash and pyrolysis temperature for SSA. This work provided new insights for understanding pyrolysis process of biomass.

Evaluating ASD in children through automatic analysis of paintings

Autism spectrum disorder (ASD) is a hereditary neurodevelopmental disorder affecting individuals, families, and societies worldwide. Screening for ASD relies on specialized medical resources, and current machine learning-based screening methods depend on expensive professional devices and algorithms. Therefore, there is a critical need to develop accessible and easily implementable methods for ASD assessment. In this study, we are committed to finding such an ASD screening and rehabilitation assessment solution based on children’s drawings. From an ASD painting database, 375 drawings from children with ASD and 160 drawings from typically developing children were selected, and a series of image signal processing algorithms based on typical characteristics of children with ASD were designed to extract features from images. The effectiveness of extracted features was evaluated through statistical methods, and they were then classified using a support vector machine (SVM) and XGBoost (eXtreme Gradient Boosting). In 5-fold cross-validation, the SVM achieved a recall of 94.93%, a precision of 86.40%, an accuracy of 85.98%, and an AUC of 90.90%, while the XGBoost achieved a recall of 96.27%, a precision of 93.78%, an accuracy of 92.90%, and an AUC of 98.00%. This efficacy persists at a high level even during additional validation on a set of newly collected paintings. Not only did the performance surpass that of participated human experts, but the high recall rate, as well as its affordability, manageability, and ease of implementation, indicates potentiality in wide screening and rehabilitation assessment. All analysis code is public at GitHub: dishangti/ASD-Painting-Pub.

Stochastic hazard assessment framework of landslide blocking river by depth-integrated continuum method and random field theory

AbstractLandslide-induced barrier dams pose a threat to the safety of humans, livestock and nearby infrastructures. The efficient assessment of landslide blocking river is crucial for disaster prevention and mitigation solutions. This study proposes a novel stochastic assessment framework to evaluate the landslide blocking river through the prediction of their deposition depths and considering the heterogeneity of shear strength parameters on the potential sliding surface. The depth-integrated continuum method (DICM) is used to simulate the landslide runout process. Using an enhanced Karhunen-Loève expansion (KLE) method, the spatial variations in soil's shear strength parameters are modeled by random fields to incorporate the effects of soil's spatial heterogeneity on the landslide deposition pattern. Subsequently, the multi-response surrogate model is constructed to relate the random field variables to the deposition depths based on extreme gradient boosting (XGBoost). To improve the performance of the surrogate model, principal component analysis (PCA) and sliced inverse regression (SIR) methods are employed for the dimension reduction of output and input variables, respectively. Furthermore, the algorithm for river blockage identification is developed to search for the deposition ridges. To demonstrate the capability of the stochastic assessment framework, an example of the first Baige landslide in Tibet, China is simulated, and the affected region and deposition depths of the landslide are predicted to calculate the probability of river damming. The presented methodology provides a practical means for improving the landslide blocking river prediction and new insights for early warning and risk mitigation.

Deciphering the impact of cascade reservoirs on nitrogen transport and nitrate transformation: Insights from multiple isotope analysis and machine learning

Construction of cascade reservoirs has altered nutrient dynamics and biogeochemical cycles, thereby influencing the composition and productivity of river ecosystems. The Lancang River (LCR), characterized by its cascade reservoir system, presents uncertainties in nitrogen transport and nitrate transformation mechanisms. Herein, we conducted monthly monitoring of hydrochemistry and multiple stable isotopes (δ15N-NO3-, δ18O-NO3-, δ18O-H2O, δD-H2O) throughout 2019 in both the natural river reach (NRR) and cascade reservoirs reach (CRR) of the LCR. Through the monthly detection of nitrogen forms and runoff in the import (M2) and export (M9) section, the average annual retention ratios for Total nitrogen (TN), Nitrate nitrogen (NO3--N), Particulate Nitrogen (PN) and Ammonium Nitrogen (NH4+-N) were about -35%, -53%, 48% and -65%, respectively. The retention rates were positively correlated with hydraulic retention time and negatively correlated with reservoir age, especially in the flood season. Compared to the NRR, the reservoir had significantly affected the nitrogen transport characteristics, especially for the large reservoirs (like Xiaowan and Nuozhadu), which enhanced phytoplankton uptake of NO3--N to form PN capabilities in the lentic environment and subsequently to precipitate or intercept it at the reservoir. This led to the overall decreasing trend of TN and PN concentrations along the CRR. The Bayesian stable isotope model quantified NO3--N sources from the NRR to the CRR. During this transition, soil nitrogen (SN) ratios decreased from 69.3% to 61.8%, while Manure & sewage (M&S) increased from 24.0% to 31.3%. Anthropogenic and natural factors, including urban sewage discharge, population density, and precipitation, were selected as key predictor variables. The eXtreme Gradient Boosting (XGBoost) model exhibited superior predictive performance for NO3--N concentrations, achieving an R2 of 0.70. These findings deepen our understanding of the impact of reservoirs on river ecology.

Multimodal radiomics and deep learning models for predicting early femoral head deformity in LCPD

PurposeTo develop a predictive model combining clinical, radiomic, and deep learning features based on X-ray and MRI to identify risk factors for early femoral head deformity in Legg-Calvé-Perthes disease (LCPD). MethodsThis study involved 152 patients diagnosed with early unilateral LCPD across two centers between January 2013 and December 2023, and included an independent external validation set to assess generalizability. Four machine learning methods, including logistic regression (LR), random forest (RF), support vector machine (SVM), and extreme gradient boosting (XGBoost), were employed to develop radiomics deep learning signatures. The clinical-radiomics model (Clinic + Rad), clinical-deep learning model (Clinic + DL), and clinical-radiomics-deep learning model (Clinic + Rad + DL) were developed by integrating radiomics deep learning signatures with clinical variables. The best model, integrated into a nomogram for clinical application, was evaluated using the area under the receiver operating characteristic curve (AUC). ResultsAmong the four machine learning methods, XGBoost demonstrated superior performance in our patient dataset: radiomic (Rad) model (AUC, 0.786) and deep learning (DL) model (AUC, 0.803). Clinical variables such as age at onset and JIC classification were associated with early femoral head deformity (p < 0.05). The combined model incorporating clinical, radiomic, and deep learning signatures demonstrated better predictive ability (AUC, 0.853). The nomogram can assist clinicians in effectively assessing the risk of early femoral head deformity. ConclusionThe Clinic + Rad + DL integrated model may be beneficial for prognostic assessment of early LCPD femoral head deformity, which is crucial for tailoring personalized treatment strategies for individual patients.