Search Results (5,077)

In predicting slope stability, updating datasets with new cases necessitates retraining traditional machine learning models, consuming substantial time and resources. This paper introduces the Incremental Learning Bayesian (ILB) model, combining incremental learning theory with the naive Bayesian model, to address this issue. Key slope parameters—height; slope angle; unit weight; cohesion; internal friction angle; and pore water ratio—are used as predictive indicators. A dataset of 242 slope cases from existing literature is compiled for training and evaluation. The ILB model’s performance is assessed using accuracy, area under the ROC curve (AUC), generalization ability, and computation time and compared to four common batch learning models: Random Forest (RF), Gradient Boosting Machine (GBM), Support Vector Machine (SVM), and Multi-Layer Perceptron (MLP). Variable importance and partial dependence plots are used to explore the relationship between prediction results and parameters. Validation is performed with real slope cases from the Lala Copper Mine in Sichuan Province, China. Results show that (1) The ILB model’s accuracy and AUC improve as the dataset grows. (2) The ILB model outperforms GBM, SVM, and MLP in accuracy and AUC, similar to RF. (3) It demonstrates superior generalization and lower computation time than batch learning models. (4) Internal friction angle, slope angle, and pore water ratio are the most important predictors. Full article

Wheat (Triticum aestivum L.) is one of the world’s primary food crops, and timely and accurate yield prediction is essential for ensuring food security. There has been a growing use of remote sensing, climate data, and their combination to estimate yields, but the optimal indices and time window for wheat yield prediction in arid regions remain unclear. This study was conducted to (1) assess the performance of widely recognized remote sensing indices to predict wheat yield at different growth stages, (2) evaluate the predictive accuracy of different yield predictive machine learning models, (3) determine the appropriate growth period for wheat yield prediction in arid regions, and (4) evaluate the impact of climate parameters on model accuracy. The vegetation indices, widely recognized due to their proven effectiveness, used in this study include the Normalized Difference Vegetation Index (NDVI), the Enhanced Vegetation Index (EVI), and the Atmospheric Resistance Vegetation Index (ARVI). Moreover, four machine learning models, viz. Decision Trees (DTs), Random Forest (RF), Gradient Boosting (GB), and Bagging Trees (BTs), were evaluated to assess their predictive accuracy for wheat yield in the arid region. The whole wheat growth period was divided into three time windows: tillering to grain filling (December 15–March), stem elongation to grain filling (January 15–March), and heading to grain filling (February–March 15). The model was evaluated and developed in the Google Earth Engine (GEE), combining climate and remote sensing data. The results showed that the RF model with ARVI could accurately predict wheat yield at the grain filling and the maturity stages in arid regions with an R² > 0.75 and yield error of less than 10%. The grain filling stage was identified as the optimal prediction window for wheat yield in arid regions. While RF with ARVI delivered the best results, GB with EVI showed slightly lower precision but still outperformed other models. It is concluded that combining multisource data and machine learning models is a promising approach for wheat yield prediction in arid regions. Full article

Monitoring the dust retention content in grassland plants around open-pit coal mines is of significant importance for environmental pollution monitoring and the development of dust control strategies. This paper focuses on the HulunBuir grassland in the Inner Mongolia Autonomous Region, China. UAV-borne hyperspectral data and measured dust retention content in plant canopies are used as data sources. The spectral response characteristics of canopy dust retention are analyzed, and four types of optimized spectral indices are constructed, including the difference index (DI), ratio index (RI), normalized difference index (NDI), and inverse difference index (IDI). The spectral index with the highest absolute value of the correlation coefficient with the canopy dust retention is selected as the feature variable for each spectral index. In addition, machine learning methods such as the partial least squares regression (PLSR), support vector machine (SVM), and random forest (RF) methods are used to develop models for the inversion of canopy dust retention. The results show that as the dust retention content increases, the canopy reflectance in the visible wavelength initially increases and then decreases, while the reflectance in the near-infrared wavelength gradually decreases. The spectral reflectance values at different dust retention levels exhibit significant differences in the 400–420 nm, 579–698 nm, and 714–1000 nm ranges. The four types of spectral indices constructed exhibit high correlations with the canopy dust retention content, and the spectral index with the highest absolute value of the correlation coefficient is composed of near-infrared bands. The dust retention inversion model established using the RF method is more accurate than those established using the PLSR and SVM methods and yields a higher prediction accuracy. The high canopy dust retention areas are mainly distributed within 900 m of the mining area, and the dust retention gradually decreases with distance. In addition, with increasing dust retention, the fractional vegetation cover (FVC) decreases. The results of this study provide a theoretical basis and technical support for monitoring dust retention in grassland plant canopies and for dust control measures. Full article

Background/Objectives: Falls among the older adult population represent a significant public health concern, often leading to diminished quality of life and serious injuries that escalate healthcare costs, and they may even prove fatal. Accurate fall risk prediction is therefore crucial for implementing timely preventive measures. However, to date, there is no definitive metric to identify individuals with high risk of experiencing a fall. To address this, the present study proposes a novel approach that transforms biomechanical time-series data, derived from gait analysis, into visual representations to facilitate the application of deep learning (DL) methods for fall risk assessment. Methods: By leveraging convolutional neural networks (CNNs) and Siamese neural networks (SNNs), the proposed framework effectively addresses the challenges of limited datasets and delivers robust predictive capabilities. Results: Through the extraction of distinctive gait-related features and the generation of class-discriminative activation maps using Grad-CAM, the random forest (RF) machine learning (ML) model not only achieves commendable accuracy (83.29%) but also enhances explainability. Conclusions: Ultimately, this study underscores the potential of advanced computational tools and machine learning algorithms to improve fall risk prediction, reduce healthcare burdens, and promote greater independence and well-being among the older adults. Full article

Accurate prediction of fuel consumption is critical for achieving efficient and low-carbon ship operations. However, the variability of the marine environment introduces significant challenges, as it leads to dynamic changes in monitoring data, complicating real-time and precise fuel consumption prediction. To address this issue, the authors proposed an incremental learning-based prediction framework to enhance adaptability to temporal dependencies in fuel consumption data. The framework dynamically adjusts a dual adaption mechanism for input features and target labels while incorporating rolling retraining to enable continuous model updates. The effectiveness of the proposed approach was validated using a real-world dataset from an LPG carrier, where it was benchmarked against conventional machine learning models, including Random Forest (RF), Linear Regression (LR), Support Vector Machine (SVM), and Multi-Layer Perceptron (MLP). Experimental results demonstrate that the proposed approach could significantly improve prediction accuracy in both offline and online scenarios. In offline mode, the proposed framework improves the R² of various machine learning models by at least 21.97%. In online mode, the proposed method increases R² by at least 17.97%. This work provides a new solution for real-time fuel consumption prediction in dynamic marine environments. Full article

China has a vast land area, with mountains accounting for 1/3 of the country’s land area. Flooding in these areas can cause significant damage to human life and property. Therefore, rainstorms and flood hazards in Huangshan City should be accurately assessed and effectively managed to improve urban resilience, promote green and low-carbon development, and ensure socio-economic stability. Through the Random Forest (RF) algorithm and the Soil Conservation Service (SCS) model, this study aimed to assess and demarcate rainstorm and flood hazard risks in Huangshan City. Specifically, Driving forces-Pressure-State-Impact-Response (DPSIR)’s framework was applied to examine the main influencing factors. Subsequently, the RF algorithm was employed to select 11 major indicators and establish a comprehensive risk assessment model integrating four factors: hazard, exposure, vulnerability, and adaptive capacity. Additionally, a flood hazard risk zoning map of Huangshan City was generated by combining the SCS model with a Geographic Information System (GIS)-based spatial analysis. The assessment results reveal significant spatial heterogeneity in rainstorm and flood risks, with higher risks concentrated in low-lying areas and urban fringes. In addition, precipitation during the flood season and economic losses were identified as key contributors to flood risk. Furthermore, flood risks in certain areas have intensified with ongoing urbanization. The evaluation model was validated by the 7 July 2020 flood event, suggesting that Huangshan District, Huizhou District, and northern Shexian County suffered the most severe economic losses. This confirms the reliability of the model. Finally, targeted flood disaster prevention and mitigation strategies were proposed for Huangshan City, particularly in the context of carbon neutrality and green urbanization, providing decision-making support for disaster prevention and emergency management. These recommendations will contribute to enhancing the city’s disaster resilience and promoting sustainable urban development. Full article

The estimation of soil organic matter (SOM) content is essential for understanding the chemical, physical, and biological functions of soil. It is also an important attribute reflecting the quality of black soil. In this study, machine learning algorithms of support vector machine (SVM), neural network (NN), decision tree (DT), random forest (RF), extreme gradient boosting machine (GBM), and generalized linear model (GLM) were used to study the accurate prediction model of SOM in Tieling County, Tieling City, Liaoning Province, China. The models were trained by using 1554 surface soil samples and 19 auxiliary variables. Recursive feature elimination was used as a feature selection method to identify effective variables. The results showed that Normalized Difference Vegetation Index (NDVI) and elevation were key auxiliary variables. Based on 10-fold cross-validation, the RF model had the highest prediction accuracy. In terms of accuracy, the coefficient of determination of RF was 0.77, and the root mean square error was 2.85. The average soil organic matter content was 20.15 g/kg. The spatial distribution of SOM shows that higher content is concentrated in the east and west, while lower content is found in the middle. The SOM content of cultivated land was lower than that of forest land. Full article

The growing sophistication of cyber threats necessitates robust and interpretable intrusion detection systems (IDS) to safeguard network security. While machine learning models such as Decision Tree (DT), Random Forest (RF), k-Nearest Neighbors (K-NN), and XGBoost demonstrate high effectiveness in detecting malicious activities, their interpretability decreases as their complexity and accuracy increase, posing challenges for critical cybersecurity applications. Local Interpretable Model-agnostic Explanations (LIME) is widely used to address this limitation; however, its reliance on normal distribution for perturbations often fails to capture the non-linear and imbalanced characteristics of datasets like CIC-IDS-2018. To address these challenges, we propose a modified LIME perturbation strategy using Weibull, Gamma, Beta, and Pareto distributions to better capture the characteristics of network traffic data. Our methodology improves the stability of different ML models trained on CIC-IDS datasets, enabling more meaningful and reliable explanations of model predictions. The proposed modifications allow for an increase in explanation fidelity by up to 78% compared to the default Gaussian approach. Pareto-based perturbations provide the best results. Among all distributions tested, Pareto consistently yielded the highest explanation fidelity and stability, particularly for K-NN ( = 0.9971, S = 0.9907) and DT ( = 0.9267, S = 0.9797). This indicates that heavy-tailed distributions fit well with real-world network traffic patterns, reducing the variance in attribute importance explanations and making them more robust. Full article

In modern agriculture, timely and accurate crop yield information is crucial for optimising agricultural production management and resource allocation. This study focused on improving the prediction accuracy of pear yields. Taking Alar City, Xinjiang, China as the research area, a variety of data including leaf area index (LAI), soil moisture (SM) and remote sensing data were collected, covering four key periods of pear growth. Three advanced algorithms, Partial Least Squares Regression (PLSR), Support Vector Regression (SVR) and Random Forest (RF), were used to construct the regression models of LAI and vegetation index in four key periods using Sentinel-2 satellite remote sensing data. The results showed that the RF algorithm provided the best results when inverting the LAI. The coefficients of determination (R²) were 0.73, 0.72, 0.76, and 0.77 for the four periods, respectively, and the root-mean-square errors (RMSE) were 0.21 m²/m², 0.24 m²/m², 0.18 m²/m², and 0.16 m²/m², respectively. Therefore, the RF algorithm was selected as the preferred method for LAI inversion in this study. Subsequently, the study further explored the potential of data assimilation techniques in enhancing the accuracy of pear yield simulation. LAI and SM were incorporated into the World Food Studies (WOFOST) crop growth model by four assimilation algorithms, namely, the Four-Dimensional Variational Approach (4D-Var), Particle Swarm Optimisation (PSO) algorithm, Ensemble Kalman Filter (EnKF), and Particle Filter (PF) in separate and joint assimilation, respectively. The experimental results showed that the assimilated model significantly improved the accuracy of yield prediction compared to the unassimilated model. In particular, the EnKF algorithm provided the highest accuracy in yield estimation with R² of 0.82, 0.79 and RMSE of 1056 kg/ha and 1385 kg/ha when LAI alone and SM alone were assimilated, whereas 4D-Var performed the best when LAI and SM were jointly assimilated, with R² as high as 0.88, and the RMSE reduced to 923 kg/ha. In addition, it was found that assimilating LAI outperformed assimilating SM when assimilating one variable, whereas joint assimilation of LAI and SM further enhanced the predictive performance beyond that of assimilating one variable alone. In summary, the present study demonstrated great potential to provide strong support for accurate prediction of pear yield by effectively integrating LAI and SM into crop growth models through data assimilation. Full article

The prediction of remaining useful life (RUL) of lithium–ion batteries is key to the reliability assessment of batteries and affects safe application of batteries. This article introduces a CEEMDAN-RF-MHA-ED-LSTM method. Using CEEMDAN, the battery capacity data were decomposed to obtain intrinsic mode functions (IMFs), and the weight of each IMF was obtained via the random forest (RF) algorithm. The LSTM neural network was used, the encoder–decoder (ED) structure was introduced, the multi-head attention (MHA) mechanism was used to construct a network model, and the particle swarm optimization (PSO) algorithm was used to optimize the model parameters. Each IMF was input into the model, and the obtained forecast results were weighted and reconstructed to obtain the final forecast data. This method was validated on the battery dataset released by NASA. Compared with that of the single LSTM model, the mean absolute error of the proposed method decreases by 74%, 62%, 71%, and 55% on the No. 05, 06, 07, and 18th battery datasets, respectively. The root mean square error decreased by 72%, 59%, 70%, and 54%, and the mean absolute percent error decreased by 75%, 65%, 71%, and 58%, respectively. This method can accurately predict battery RUL. Full article

Background: Early safety outcomes following transcatheter aortic valve implantation (TAVI) for severe aortic stenosis are critical for patient prognosis. Accurate prediction of adverse events can enhance patient management and improve outcomes. Aim: This study aimed to develop a machine learning model to predict early safety outcomes in patients with severe aortic stenosis undergoing TAVI. Methods: We conducted a retrospective single-centre study involving 224 patients with severe aortic stenosis who underwent TAVI. Seventy-seven clinical and biochemical variables were collected for analysis. To handle unbalanced classification problems, an adaptive synthetic (ADASYN) sampling approach was used. A fined-tuned random forest (RF) machine learning model was developed to predict early safety outcomes, defined as all-cause mortality, stroke, life-threatening bleeding, acute kidney injury (stage 2 or 3), coronary artery obstruction requiring intervention, major vascular complications, and valve-related dysfunction requiring repeat procedures. Shapley Additive Explanations (SHAPs) were used to explain the output of the machine learning model by attributing each variable’s contribution to the final prediction of early safety outcomes. Results: The random forest model identified left femoral artery diameter and aortic valve calcification volume as the most influential predictors of early safety outcomes. SHAPs analysis demonstrated that smaller left femoral artery diameter and higher aortic valve calcification volume were associated with poorer early safety prognoses. Conclusions: The machine learning model highlights of early safety outcomes after TAVI. These findings suggest that incorporating these variables into pre-procedural assessments may improve risk stratification and inform clinical decision-making to enhance patient care. Full article

To effectively mitigate the hazards posed by the blast waves of rock mass caving on closed walls during the mining process, a stability prediction method based on a random forest (RF) algorithm is proposed, which is designed to automatically identify key parameters. A machine learning model is developed using the algorithm, and its performance is evaluated through accuracy, precision, recall, and F1-score metrics. The probabilistic model of the objective function is constructed using the grid search hyperparameter optimization method, allowing for the selection of the most favorable hyperparameters for evaluation. The initial prediction accuracy of the RF algorithm model is 94.6%, indicating a strong predictive capability. Further adjustments to the base classifier, maximum depth, minimum number of leaves, and minimum number of samples enhance the model’s performance, resulting in an improved prediction accuracy of 95.9%. Finally, the optimized model is applied to predict the stability of the closed walls in the actual project, and the results are consistent with the on-site situation. This demonstrates that the random forest-based stability prediction model effectively forecasts the stability of closed walls in the actual project. Full article

In this study, temperature estimation was achieved by utilizing artificial neural network (ANN) and machine learning models (linear model, support vector machine, K-nearest neighbor, random forest) to assist with sustainable environmental planning and climate change adaptation solutions. The research compared monthly humidity, wind speed, precipitation, and temperature data of the Istanbul province from 1950 to 2023. Estimates with 96% accuracy were achieved with the ANN model, and amongst the machine learning models, the random forest (RF) model demonstrated the highest performance. Generalization capability of the models was enhanced by the k-fold cross-validation method. The analysis found input variables (humidity, wind, precipitation) to be negatively associated with temperature. The current results show that the application of artificial intelligence/machine learning techniques is a useful instrument in the context of sustainable climate monitoring and temperature estimation. This study achieves sustainability targets through certain reliable methodologies for climate change evaluation, sustainable energy design, and agricultural adaptation plans. The methodology is transferable to other regional climate analyses and has the potential to underpin evidence-based, decision making for sustainable development and climate resilience. Full article

Historically known for its tin mining industry, Phuket Island has undergone significant transformation into a global tourism hub. This study aims at analyzing the evolutionary dynamics of Phuket Island from the years 1987 to 2024. We integrate Landsat satellite images and sophisticated analytical methods to assess the effects of tourism and economic policies on changes in land use and land cover using Google Earth Engine (GEE) for cloud-based data processing and Random Forest (RF) models for classification, and the Urban Expansion Intensity Index (UEII) and Shannon Entropy metrics for measuring the intensity of urban expansion and diversity, respectively. The results show that there has been a dynamic change in the patterns of land use which was brought about by the economic and environmental forces. Some of the major events that have had a great effect on Phuket’s landscape include the 1997 Asian Financial Crisis, the 2004 Indian Ocean Tsunami, and the COVID-19 pandemic; this highlights how the island is fragile and can be affected easily by events happening around the world. This work reveals a dramatic reduction in forest and mangrove cover, which calls for increased conservation measures to prevent the loss of biodiversity and to preserve the natural balance. Full article

This study examines the efficacy of Random Forest and XGBoost classifiers in conjunction with three upsampling techniques—SMOTE, ADASYN, and Gaussian noise upsampling (GNUS)—across datasets with varying class imbalance levels, ranging from moderate to extreme (15% to 1% churn rate). Employing metrics such as F1 score, ROC AUC, PR AUC, Matthews Correlation Coefficient (MCC), and Cohen’s Kappa, this research provides a comprehensive evaluation of classifier performance under different imbalance scenarios, focusing on applications in the telecommunications domain. The findings highlight that tuned XGBoost paired with SMOTE (Tuned_XGB_SMOTE) consistently achieves the highest F1 score and robust performance across all imbalance levels. SMOTE emerged as the most effective upsampling method, particularly when used with XGBoost, whereas Random Forest performed poorly under severe imbalance. ADASYN showed moderate effectiveness with XGBoost but underperformed with Random Forest, and GNUS produced inconsistent results. This study underscores the impact of data imbalance, with MCC, Kappa, and F1 scores fluctuating significantly, whereas ROC AUC and PR AUC remained relatively stable. Moreover, rigorous statistical analyses employing the Friedman test and Nemenyi post hoc comparisons confirmed that the observed improvements in F1 score, PR-AUC, Kappa, and MCC were statistically significant (p < 0.05), with Tuned_XGB_SMOTE significantly outperforming Tuned_RF_GNUS. While differences in ROC-AUC were not significant, the consistency of these results across multiple performance metrics underscores the reliability of our framework, offering a statistically validated and attractive solution for model selection in imbalanced classification scenarios. Full article