Ensemble Model Research Articles

Application of CT-based foundational AI and radiomics models for prediction of outcome for non-small-cell lung cancer patients treated on the NRG/RTOG 0617 clinical trial

Abstract Objectives To apply CT-based foundational artificial intelligence (AI) and radiomics models for predicting overall survival (OS) for patients with locally advanced non-small cell lung cancer (NSCLC). Methods Data for 449 patients retrospectively treated on the NRG Oncology/Radiation Therapy Oncology Group (RTOG) 0617 clinical trial were analyzed. Foundational AI, radiomics and clinical features were evaluated using univariate cox regression and correlational analyses to determine independent predictors of survival. Several models were fit using these predictors and model performance was evaluated using nested cross-validation and unseen independent test datasets via area under receiver-operator-characteristic curves, AUC’s. Results For all patients, the combined foundational AI and clinical models achieved AUCs of 0.67 for the Random Forest (RF) model. The combined radiomics and clinical models achieved RF AUCs of 0.66. In the low-dose arm, foundational AI alone achieved AUC of 0.67, while AUC for the ensemble radiomics and clinical models was 0.65 for the support vector machine (SVM). In the high-dose arm, AUC values were 0.67 for combined radiomics and clinical models and 0.66 for the foundational AI model. Conclusions This study demonstrated encouraging results for application of foundational AI and radiomics models for prediction of outcomes. More research is warranted to understand the value of ensemble models toward improving performance via complementary information. Advances in knowledge Using foundational AI and radiomics-based models we were able to identify significant signatures of outcomes for NSCLC patients retrospectively treated on a national cooperative group clinical trial. Associated models will be important for application toward prospective patients.

Performance comparison of several explainable hybrid ensemble models for predicting carbonation depth in fly ash concrete

The carbonation of fly ash concrete critically impacts the lifespan of structures, necessitating precise prediction of carbonation depth for the construction industry. This study establishes an original database comprising 883 cases, which are divided into training and testing sets in a 4:1 ratio. The Sand Cat Swarm Algorithm (SCSO) was developed to optimize the hyperparameters of three ensemble models: Gradient Boosting Decision Trees (GBDT), Light Gradient Boosting Machine (LGBM), and Categorical Boosting (CatBoost), resulting in the development of three hybrid ensemble models: SCSO-GBDT, SCSO-LGBM, and SCSO-CatBoost. Five classic models were included in the comparison, and all models used five-fold cross validation. Models’ performance was rigorously evaluated using Correlation coefficient (R2), Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Variance Accounted For (VAF), with VIsekriterijumska optimizacija i KOmpromisno Resenje (VIKOR) method and Taylor diagrams utilized for optimal model selection. The SCSO-CatBoost model demonstrated superior performance, with R2 = 0.9657, MAE = 2.0062, RMSE = 8.2147, and VAF = 96.6177. Shapley additive explanations (SHAP) analysis of the SCSO-CatBoost model identified time of exposure as the most significant factor influencing carbonation depth, followed by fly ash content and carbon dioxide concentration. To facilitate practical application by non-algorithm engineers, an intelligent program was developed, allowing for straightforward testing with the three hybrid ensemble models. This study presents three precise models for predicting the carbonation depth of fly ash concrete. These models serve as valuable tools for estimating the service life of concrete structures and can be utilized as simulation instruments in durability engineering.

Catalyzing IVF outcome prediction: exploring advanced machine learning paradigms for enhanced success rate prognostication

This study addresses the research problem of enhancing In-Vitro Fertilization (IVF) success rate prediction by integrating advanced machine learning paradigms with gynecological expertise. The methodology involves the analysis of comprehensive datasets from 2017 to 2018 and 2010–2016. Machine learning models, including Logistic Regression, Gaussian NB, SVM, MLP, KNN, and ensemble models like Random Forest, AdaBoost, Logit Boost, RUS Boost, and RSM, were employed. Key findings reveal the significance of patient demographics, infertility factors, and treatment protocols in IVF success prediction. Notably, ensemble learning methods demonstrated high accuracy, with Logit Boost achieving an accuracy of 96.35%. The implications of this research span clinical decision support, patient counseling, and data preprocessing techniques, highlighting the potential for personalized IVF treatments and continuous monitoring. The study underscores the importance of collaboration between gynecologists and data scientists to optimize IVF outcomes. Prospective studies and external validation are suggested as future directions, promising to further revolutionize fertility treatments and offer hope to couples facing infertility challenges.

Machine Learning-Based Prediction for Incident Hypertension Based on Regular Health Checkup Data: Derivation and Validation in 2 Independent Nationwide Cohorts in South Korea and Japan.

Worldwide, cardiovascular diseases are the primary cause of death, with hypertension as a key contributor. In 2019, cardiovascular diseases led to 17.9 million deaths, predicted to reach 23 million by 2030. This study presents a new method to predict hypertension using demographic data, using 6 machine learning models for enhanced reliability and applicability. The goal is to harness artificial intelligence for early and accurate hypertension diagnosis across diverse populations. Data from 2 national cohort studies, National Health Insurance Service-National Sample Cohort (South Korea, n=244,814), conducted between 2002 and 2013 were used to train and test machine learning models designed to anticipate incident hypertension within 5 years of a health checkup involving those aged ≥20 years, and Japanese Medical Data Center cohort (Japan, n=1,296,649) were used for extra validation. An ensemble from 6 diverse machine learning models was used to identify the 5 most salient features contributing to hypertension by presenting a feature importance analysis to confirm the contribution of each future. The Adaptive Boosting and logistic regression ensemble showed superior balanced accuracy (0.812, sensitivity 0.806, specificity 0.818, and area under the receiver operating characteristic curve 0.901). The 5 key hypertension indicators were age, diastolic blood pressure, BMI, systolic blood pressure, and fasting blood glucose. The Japanese Medical Data Center cohort dataset (extra validation set) corroborated these findings (balanced accuracy 0.741 and area under the receiver operating characteristic curve 0.824). The ensemble model was integrated into a public web portal for predicting hypertension onset based on health checkup data. Comparative evaluation of our machine learning models against classical statistical models across 2 distinct studies emphasized the former's enhanced stability, generalizability, and reproducibility in predicting hypertension onset.

FuNet-40: fundus disease/abnormality classification using ensemble of fine-tuned pretrained convolution models

FuNet-40: fundus disease/abnormality classification using ensemble of fine-tuned pretrained convolution models

Evidence of repeated zoonotic pathogen spillover events at ecological boundaries

Anthropogenic modifications to the landscape have altered several ecological processes worldwide, creating new ecological boundaries at the human/wildlife interface. Outbreaks of zoonotic pathogens often occur at these ecological boundaries, but the mechanisms behind new emergences remain drastically understudied. Here, we test for the influence of two types of ecosystem boundaries on spillover risk: (1) biotic transition zones such as species range edges and transitions between ecoregions and (2) land use transition zones where wild landscapes occur in close proximity to heavily impacted areas of high human population density. Using ebolavirus as a model system and an ensemble machine learning modeling framework, we investigated the role of likely reservoir (bats) and accidental host (primates) range edges and patterns of land use (defined using SEDAC categories) on past spillover events. Our results show that overlapping species range edges and heightened habitat diversity increase ebolavirus outbreaks risk. Moreover, we show that gradual transition zones, represent by high proportion of rangelands, acts as a buffer to reduces outbreak risks. With increasing landscape changes worldwide, we provide novel ecological and evolutionary insights into our understanding of zoonotic pathogen emergence and highlight the risk of aggressively developing ecological boundaries.

Unveiling the nonlinear drivers of urban land resources on carbon emissions: The mediating role of industrial upgrading and technological innovation

As cities continue to develop, land resources play a pivotal role in the pursuit of sustainable urban development. However, the impact and mechanisms of land resources on reducing emissions remain unclear. Therefore, our study employs an ensemble learning model within a Monte Carlo Simulation framework to explore the nonlinear relationship between land resources and carbon emissions. Additionally, we utilize a fixed effects model to identify the mechanisms of impact. Specifically, industrial and residential land uses are identified as primary drivers of carbon emissions, contributing 17.04 % and 4.75 % of the total variance in emissions, respectively. In contrast, a standard deviation increase in land price leads to a reduction of 4.65 tons in per capita carbon emissions. In addition, industrial upgrading and innovation emerge as critical mediators in promoting carbon reduction. This research contributes to sustainable urban planning and policy-making, offering insights into how strategic land use planning reduces urban carbon footprints.

Addressing catastrophic forgetting in payload parameter identification using incremental ensemble learning

Collaborative robots (cobots) are increasingly integrated into Industry 4.0 dynamic manufacturing environments that require frequent system reconfiguration due to changes in cobot paths and payloads. This necessitates fast methods for identifying payload inertial parameters to compensate the cobot controller and ensure precise and safe operation. Our prior work used Incremental Ensemble Model (IEM) to identify payload parameters, eliminating the need for an excitation path and thus removing the separate identification step. However, this approach suffers from catastrophic forgetting. This paper introduces a novel incremental ensemble learning method that addresses the problem of catastrophic forgetting by adding a new weak learner to the ensemble model for each new training bag. Moreover, it proposes a new classification model that assists the ensemble model in identifying which weak learner provides the most accurate estimation for new input data. The proposed method incrementally updates the identification model while the cobot navigates any task path, maintaining accuracy on old weak learner even after updating with new data. Validation performed on the Franka Emika cobot showcases the model’s superior accuracy and adaptability, effectively eliminating the problem of catastrophic forgetting.

A fast prediction method of fatigue life for crane structure based on Stacking ensemble learning model

To quickly obtain the fatigue life of cranes in service, the metal structure that determines the crane life is anchored. Meanwhile, the fast prediction method of fatigue life of crane metal structures based on the Stacking ensemble learning model is proposed. Firstly, in line with the structural stress method, the global rough model of the metal structure is established by the co-simulation technology to obtain the fatigue damage regions of the structure. The local fine model is constructed by local cutting and boundary condition transplantation to determine the critical weld at the failure regions. Secondly, through weld definition, equivalent structural stress acquisition, and fatigue life calculation, the sample data set with lifting load and trolley running position as input and fatigue life cycle times as output is constructed. Then, the Stacking integrated learning model combining gradient boosting, ridge regression, Extra Trees, and linear is built. On this basis, combined with the Miner theory, the rapid prediction of crane fatigue life is realized. Finally, the proposed method is applied to the QD40t × 22.5 m × 9 m general bridge crane. The results show that the life sample set constructed by the structural stress method is more accurate and reasonable than the nominal, hot spot, and fracture mechanics methods. The life prediction results of the Stacking integration model were improved by 6.3 to 49.2% compared to the single model. The method has theoretical and practical significance in reducing accidents and ensuring the safe operation of cranes.

Enhancing discharge prediction over Type-A piano key weirs: An innovative machine learning approach

Piano key weirs (PKWs) are an increasingly popular hydraulic structure due to their higher discharge capacity than linear weirs. Accurately predicting the discharge of PKWs is essential for appropriate design and operation. This study utilized eight Machine Learning algorithms, including non-ensemble and ensemble models to predict the discharge of type-A PKWs. Multiple-Linear-Regression (MLR), Support-Vector-Machine (SVM), Gene-Expression-Programming (GEP), and Artificial-Neural-Network (ANN) were adopted as non-ensemble models. While the ensemble models comprised Random-Forest (RF), Adaptive-Boosting (AdaBoost), Extreme-Gradient-Boosting (XGBoost), and Categorical-Boosting (CatBoost). A total of 476 experimental datasets were collected from previous research considering three critical dimensionless input parameters: PKW key widths, PKW height, and total upstream head. The models were trained on 70 % of the dataset and tested on the remaining 30 %. The hyperparameters of the models were optimized using the Bayesian Optimization technique, with 5-fold cross-validation ensuring high performance. Comprehensive analyses, including visual and quantitative methods, were employed to validate model effectiveness. CatBoost model consistently outperformed the other models, achieving the highest Determination-coefficient (R2 = 0.998) and lowest Root-Mean-Squared-Error (RMSE = 0.002), highlighting its ability to handle complex data patterns and its superior optimization process. XGBoost follows closely behind, showing strong generalization, while ANN and RF perform well, but it's a slight increase in error metrics. The study also incorporated Shapley-Additive-exPlanations (SHAP) and Partial-Dependence-Plot (PDP) analyses, revealing that the total upstream head variable had the most significant impact on the discharge predictions. An interactive Graphical-User-Interface was developed to facilitate practical applications, enabling engineers to predict discharge quickly and economically.

Improved machine learning models with a similarity-based approach for remaining useful life prediction

Abstract Cost efficiency is a critical factor in the competitive aviation sector. These efficiency factors force airline operators to develop new approaches in their organizations. Predictive maintenance helps to build scheduling maintenance programs for airline operators or MROs. Scheduled maintenance programs benefit cost efficiency in the aviation sector. Predictive maintenance methods predict the failure time of any equipment. Predictions can be made by analyzing the sensor values from equipment. In this paper, we predicted the remaining useful life (RUL) of turbofan engines using machine learning models and a similarity-based approach. Sensor datasets from the Prognostics Data Repository of NASA, called CMAPPS, were utilized. Using the FD0002 sub-dataset, a health index (HI) was created, and models were trained. Once the models were trained, train and test HIs were estimated. The predicted test HI was matched with the predicted train HI based on a similarity-based approach, and then a RUL prediction was made. The results obtained were compared with the actual results to calculate the accuracy, and the algorithm that resulted in the maximum accuracy was identified. We selected six machine learning algorithms and also created an ensemble model by averaging the predictions of six machine learning algorithms for comparing prediction accuracy. The different algorithms were compared to obtain the prediction model with the closest prediction of remaining useful lifecycle in terms of the number of life cycles. This experiment showed us the effect of the similarity-based approach on the basic version of machine learning models for RUL prediction.

Embankment Breach Simulation and Inundation Mapping Leveraging High-Performance Computing for Enhanced Flood Risk Prediction and Assessment

Abstract. Embankment breaches represent significant hazards to communities and infrastructure, precipitating catastrophic flood occurrences. The precise prediction of floods and understanding the scope of inundation stemming from embankment failures are imperative for effective disaster preparedness and response. This research delves into a case study on simulating embankment breaches to evaluate the extent of flooding. Leveraging advanced hydrodynamic models validated through high-performance computing (HPC) systems, and integrating real-time data assimilation, we aim to improve accuracy in flood forecasting. The study endeavours to bolster flood risk management by furnishing detailed inundation maps and insights into embankment breach dynamics, thereby facilitating enhanced preparedness and response strategies. Our findings reveal that simulations conducted on multicore processors offer superior performance compared to single-core setups, yielding enhanced result accuracy and providing administrators with increased lead time. It unlocks high-resolution simulations for intricate basin details, explores a wider range of flood scenarios quickly, and allows for efficient ensemble modelling to assess model uncertainty. Through HPC utilization, we can harness high-resolution digital elevation models (DEMs) for 2D-hydrodynamic modelling, enabling rapid assessment of water spread resulting from embankment breaches within a mere 20-minute timeframe, a significant improvement from the previous 3–4 hours duration.

Conductivity Prediction Method of Solid Electrolyte Materials Based on Pearson Coefficient Method and Ensemble Learning

In order to screen solid electrolytic materials with high ionic conductivity more quickly and accurately, a conductivity prediction method for solid electrolyte materials based on logistic regression model and random forest regression model is constructed. The method first selects 20 characteristic descriptors related to ionic conductivity from material library according to four dimensions: structural stability, metal stability, electronic conductivity and oxidative decomposition stability. The enumeration method and feature selection method are used to combine different features. Then, the dimension reduction of the feature combinations is carried out by Pearson coefficient method, so as to screen the optimal feature combinations. Finally, according to the optimal feature combination and data subset, the constructed random forest ensemble learning model is utilized to predict the ionic conductivity of other solid-state electrolyte materials in the material library, in order to find the solid-state electrolyte materials that can meet the high conductivity. The results indicate that as the number of features in the combination increases, the prediction performance of the model shows a trend of first increasing and then decreasing. When the feature descriptor in the combination is 7, the maximum AUC (Accuracy) value is reached, and the optimal feature in this case contains 7 optimal feature subsets corresponding to feature descriptors, namely the standard deviation of the average adjacency number of Li atoms, the standard deviation of Li-X ion bonds, the average electronegativity of the straight path, the average width of the straight path, the average atomic volume, the average Li-Li bond, and the filling fraction of the sublattice. On the basis of the optimal feature screening, the average prediction accuracy of the logistic regression model is 87.13%, which has a high accuracy. Using the random forest regression model for prediction, the average absolute error and root mean square error obtained are only 0.237 and 0.134, and compared with classical classification methods such as KNN, SVM and Adaboost, they are smaller, which can better predict the ionic conductivity of solid electrolyte materials. This proves that the constructed method can quickly and accurately produce high ionic conductivity materials in solid electrolyte materials, which is worthy of further research and promotion.

Analyzing the Distribution Patterns of Endemic Quercus vulcanica (Boiss. et Heldr. ex) Kotschy in Türkiye Under Climate Change Using Ensemble Modeling

Quercus vulcanica (Boiss. et Heldr. ex) Kotschy (Kasnak oak), one of the 18 Quercus species naturally distributed in Anatolia, is an endemic species with a restricted distribution range. In accordance with the International Union for the Conservation of Nature (IUCN) Red List of Endangered Species classification, Quercus vulcanica is designated as a species of low risk (LC: Least Concern). However, it is predicted that the habitat of Quercus vulcanica will narrow and that the species will become endangered as a result of potential climate change scenarios in the future. The aim of this study was to estimate the temporal and spatial distribution of Quercus vulcanica in Anatolia during the LGM, as well as to examine the impact of present and future climate changes on the species. In this context, principal component analysis was applied to 19 bioclimatic variables of the Community Climate System Model Version 4 (CCSM4) climate model, with nine variables identified for use in modeling. Habitat suitability was estimated using the Biodiversity Modeling (BIOMOD) ensemble modeling method, which combines the results of nine different algorithms through the R package ‘biomod2’, applying both committee averaging and weighted average approaches. To evaluate the performance of the models, the Area Under the Curve (AUC) of Receiver Operating Characteristics (ROC), True Skill Statistics (TSS), KAPPA and Boyce Index were calculated. The contributions of the environmental variables were determined on a per-algorithm-model basis. The results of the analyses show that the bioclimatic variables that contribute the most to the distribution of the species are Bio8. The modeling results show that Quercus vulcanica is capable of occupying suitable habitat areas across the majority of Anatolia during the Last Glacial Maximum (LGM). It is anticipated that future projections will indicate a notable reduction in the extent of suitable habitat for the species, with the remaining areas confined to the vicinity of the Ilgaz Mountains, Köroğlu Mountains and Bolkar Mountains. Given the increasing destruction that Quercus vulcanica, an endemic plant, will be adversely affected by as a result of human impacts and climate change, it is of the highest importance to develop adaptation strategies with a view to protecting the species’ habitat and the sustainability of the species.

Prediction of Suitable Regions for Danxiaorchis yangii Combined with Pollinators Based on the SDM Model.

Danxiaorchis yangii, a newly discovered fully mycoheterotrophic orchid. It relies on Lysimachia alfredii and Dufourea spp. for pollination, and environmental factors closely influence the growth and distribution of these pollinators, which in turn directly affects the growth and reproduction of D. yangii. Climate change threatens the suitable habitats for these three species, emphasizing the need to understand D. yangii's response. This study comprehensively utilized the field distribution of D. yangii and related climatic data, along with future climate predictions from global models, to predict the climate suitability areas of D. yangii under two greenhouse gas emission scenarios (SSP245 and SSP370) using species distribution models (SDMs), which encompassed a random forest (RF) model. Additionally, we selected the optimal ensemble model (OEM) for Dufourea spp. and applied generalized boosted models (GBMs) and RF for L. alfredii in our predictions. The study found that precipitation of the driest quarter plays a pivotal role in determining the distribution of D. yangii, with an optimal range of 159 to 730 mm being most conducive to its growth. Comparative analysis further indicated that precipitation exerts a greater influence on D. yangii than temperature. Historically, D. yangii has been predominantly distributed across Jiangxi, Hunan, Zhejiang, and the Guangxi Zhuang Autonomous Region, with Jiangxi Province containing the largest area of highly suitable habitat, and this distribution largely overlaps with the suitable regions of its pollinators.

Efficient Flood Detection through Hybrid Machine Learning and Metaheuristic Methods using Sentinel-1

Abstract. Floods are considered among the most destructive natural disasters, requiring precise and timely management. Remote sensing, utilizing diverse satellite imagery data, enables effective monitoring and assessment of flood impacts. In this context, machine learning and deep learning methods, as effective and scalable approaches, can significantly enhance the accuracy of flood detection and management by analyzing remote sensing data, thereby playing a crucial role in mitigating flood-related risks. In this study, to flood detection using Sentinel-1 SAR data, machine learning algorithms, including Random Forest (RF) and Histogram-based Gradient Boosting Decision Tree, were employed, along with two metaheuristic algorithms, Harris Hawks Optimization (HHO) and Ant Colony Optimization (ACO), for hyperparameter optimization. Additionally, to enhance the models' ability to detect flooded pixels and improve overall performance accurately, a pre-trained VGG-16 Neural Network was used as a deep feature extractor. Finally, four ensemble flood detection models—RF-HHO, RF-ACO, HGBDT-HHO, and HGBDT-ACO—were implemented, and their performance was evaluated and compared based on statistical metrics. Based on the obtained results, all four ensemble flood detection models demonstrated excellent performance in the validation and testing phases. The overall accuracy of these models reached over 95% in the validation phase and exceeded 97% in the testing phase. However, the HGBDT-ACO model achieved the highest accuracy and the lowest error rate in detecting flood pixels, making it the best-performing model in this study. Generally, HGBDT models showed a relative advantage over RF models, as they required significantly less time for training while achieving comparable results. Therefore, they were efficient and performed better in terms of computational complexity.

Securing transactions: a hybrid dependable ensemble machine learning model using IHT-LR and grid search

Financial institutions and businesses face an ongoing challenge from fraudulent transactions, prompting the need for effective detection methods. Detecting credit card fraud is crucial for identifying and preventing unauthorized transactions. While credit card fraud incidents are relatively rare, they can result in substantial financial losses, particularly due to the high monetary value associated with fraudulent transactions. Timely detection of fraud enables investigators to take swift actions to mitigate further losses. However, the investigation process is often time-consuming, limiting the number of alerts that can be thoroughly examined each day. Therefore, the primary objective of a fraud detection model is to provide accurate alerts while minimizing false alarms and missed fraud cases. In this paper, we introduce a state-of-the-art hybrid ensemble (ENS) dependable machine learning (ML) model that intelligently combines multiple algorithms with proper weighted optimization using grid search, including decision tree (DT), random forest (RF), K-nearest neighbor (KNN), and multilayer perceptron (MLP), to enhance fraud identification. To address the data imbalance issue, we employ the instant hardness threshold (IHT) technique in conjunction with logistic regression (LR), surpassing conventional approaches. Our experiments are conducted on a publicly available credit card dataset comprising 284,807 transactions. The proposed model achieves impressive accuracy rates of 99.66%, 99.73%, 98.56%, and 99.79%, and a perfect 100% for the DT, RF, KNN, MLP and ENS models, respectively. The hybrid ensemble model outperforms existing works, establishing a new benchmark for detecting fraudulent transactions in high-frequency scenarios. The results highlight the effectiveness and reliability of our approach, demonstrating superior performance metrics and showcasing its exceptional potential for real-world fraud detection applications.

Landslide Displacement Prediction Stacking Deep Learning Algorithms: A Case Study of Shengjibao Landslide in the Three Gorges Reservoir Area of China

Computational models enable accurate, timely prediction of landslides based on the monitoring data on-site as the development of artificial intelligence technology. The most existing prediction methods focus on finding a single prediction algorithm with excellent performance or an integrated and efficient hyperparameter optimization algorithm with a highly accurate regression prediction algorithm. In order to break through the limitation of generalization of prediction models, this paper proposes an ensemble model that combines deep learning algorithms, with a stacking framework optimized with the sliding window method. Multiple deep learning algorithms are set as the first layer of the stacking framework, which is optimized with the sliding window method to avoid confusion in the time order of datasets based on time series analysis. The Shengjibao landslide in the Three Gorges Reservoir is used as a case study. First, the cumulative displacement is decomposed into a trend and a periodic term using a moving average method. A single-factor and a multi-factor superposition model based on multiple deep learning algorithms are used to predict the trend and periodic term of the displacement, respectively. Finally, the predicted values of the trend and periodic terms are added to obtain the total predicted landslide displacement. For monitoring point ZK2-3, the values of RMSE and MAPE of the total displacement prediction with the stacking model are 15.93 mm and 0.54%, and the values of RMSE and MAPE of the best-performing individual deep learning model are 20.00 mm and 0.64%. The results show that the stacking model outperforms other models by combining the advantages of each individual deep learning algorithm. This study provides a framework for integrating landslide displacement prediction models. It can serve as a reference for the geological disaster prediction and the establishment of an early warning system in the Three Gorges Reservoir Area.

Hyperspectral Inversion of Soil Cu Content in Agricultural Land Based on Continuous Wavelet Transform and Stacking Ensemble Learning

Heavy metal pollution in agricultural land poses significant threats to both the ecological environment and human health. Therefore, the rapid and accurate prediction of heavy metal content in agricultural soil is crucial for environmental protection and soil remediation. Acknowledging the limitations of traditional single linear or nonlinear machine learning models in terms of prediction accuracy, this study developed an ensemble learning model that integrates multiple linear or nonlinear learning models with a random forest (RF) model to improve both the prediction accuracy and reliability. In this study, we selected a typical copper (Cu) polluted area in the Pearl River Delta of Guangdong Province as the research site and collected Cu content data and indoor soil reflectance spectral data from 269 surface soil samples. First, the soil spectral data were preprocessed using Savitzky–Golay (SG) smoothing, multiplicative scattering correction (MSC), and continuous wavelet transform (CWT) to reduce noise interference. Next, principal components analysis (PCA) was employed to reduce the dimensionality of the preprocessed spectral data, eliminating redundant features and lowering the computational complexity. Finally, based on the dimensionality-reduced data and Cu content, we established a stacked ensemble learning model, where the base models included SVR, PLSR, BPNN, and XGBoost, with RF serving as the meta-model to estimate the soil heavy metal content. To evaluate the performance of the stacking model, we compared its prediction accuracy with that of individual models. The results indicate that, compared to the traditional machine learning models, the prediction accuracy of the stacking model was superior (R2 = 0.77; RMSE = 7.65 mg/kg; RPD = 2.29). This suggests that the integrated algorithm demonstrates a greater robustness and generalization capability. This study presents a method to improve soil heavy metal content estimation using hyperspectral technology, ensuring a robust model that supports policymakers in making informed decisions about land use, agriculture, and environmental protection.

Short-term inbound passenger flow prediction at high-speed railway stations considering the departure passenger arrival pattern

Accurate prediction of short-term inbound passenger flow at high-speed railway stations is of great significance for the refined operation of stations, the formulation of emergency plans, and the provision of intelligent services. The arrival of passengers traveling on the same train at the same station shows a similar pattern, which is called the departure passenger arrival pattern (DPAP). The short-term inbound passenger flow at the station is composed of the short-term inbound passenger flow of all waiting trains within the same period. Inspired by this, this paper develops an ensemble prediction model based on the time series decomposition modeling strategy to introduce the DPAP to the short-term inbound passenger flow prediction at stations. Firstly, we propose a new framework for studying the DPAP to calculate the fitted station short-term inbound passenger flow, which is only affected by the DPAP. During this process, we find that 7minutes is the optimal time granularity. Secondly, based on the singular spectrum analysis, we prove that the DPAP is the determining factor affecting the station short-term inbound passenger flow. Finally, we propose an ensemble prediction model that considers the DPAP to achieve short-term inbound passenger flow prediction at stations. The model consists of two parts: the deterministic and stochastic components prediction, where the former is predicted by the fitted station short-term inbound passenger flow, and the latter is achieved by the combination of historical stochastic components and weather type with the help of the Seq2Seq model based on time attention mechanism. Using real inbound passenger flow data, we compare the proposed model with 13 benchmark models and the results show that under different training and prediction steps, our model achieves optimal prediction performance, whether in all-day period and the busiest period of the station. Through further ablation experiments, it has been proven that the introduction of the DPAP effectively improves the prediction accuracy. Our model can provide scientific support for the intelligent operation of stations and the refined management of passenger flow.