How do experts classify sepsis cases for sepsis surveillance? Lessons learned from a Behavioural Artificial Intelligence Technology (BAIT) approach.

Screening for depression risk in asthma patients: Development and external validation of a machine learning-based predictive model.

Although vegetation, as a significant part of all terrestrial ecosystems, is influenced by both climatic and anthropogenic factors, identifying the relative contribution of these factors to vegetation variation remains a challenge. This article aims to assess the relative contribution of these factors to vegetation dynamics in the Gavkhuni watershed, Iran, using the residual trend analysis and machine learning algorithm, considering the Enhanced Vegetation Index (EVI) and Standardized Precipitation Evapotranspiration Index (SPEI). To achieve this goal, a time series of the EVI over the period 2001-2023 was obtained from MOD13Q1 through the Google Earth Engine platform. Also, a time series of 03, 06, 09, and 12-month SPEI was calculated using the meteorological data related to 2001-2023. Based on SPEI and EVI, the residual trend analysis was done in TerrSet 19.0.6 software, and the relative contribution of human and climatic factors to EVI dynamics was assessed. Then, considering environmental parameters besides climatic and human factors, the random forest algorithm was utilized to model the contribution of humans and climate to vegetation variation and specify the importance of variables in model efficiency. The results demonstrated that climate was responsible for the decrease in vegetation in approximately 20% of the watershed area, and human factors were the major driver for the increase in vegetation in about 38% of the Gavkhuni watershed area due to agricultural and gardening activities. According to the machine learning outcomes, climatic parameters played a significant role in vegetation decline, mostly in the northwest of the study area, and humans played an important role in vegetation increase in the west, southwest, and southeast parts of the watershed. These findings can be useful in guiding environmental policies and regional vegetation restoration projects.

Complex network modeling has been applied in sports science research; yet few studies have applied it to capture the dynamic evolution of multivariate relationships within a single session of high-intensity training. This study is the first to use complex network modeling to examine the dynamic associations among kinematic (stroke rate, stroke length), metabolic (blood lactate), and perceived exertion (Rating of Perceived Exertion, RPE) variables and 100-meter freestyle performance. The analysis was conducted on 16 adolescent swimmers during a 6 × 50-meter sprint interval training (6 × 50m SSIT) protocol. The findings showed that the overall network topology remained stable throughout the SSIT protocol, suggesting that multiple training bouts within the session collectively contributed to enhancing 100-meter freestyle performance (network density: from 42.65% to 49.17%; modularity: from 0.2 to 0.24). Nevertheless, the relative importance of individual variables shifted markedly during the training process. Specifically, the nodal centrality of swimming velocity, blood lactate, and RPE increased substantially, positioning them as central hubs mediating performance outcomes. In contrast, the influence of stroke rate progressively declined, whereas stroke length remained relatively stable. This research introduces a powerful analytical tool for the dynamic assessment of training processes and provides valuable insights into the adaptive mechanisms shaping sport-specific anaerobic capacity in competitive swimmers.

ABSTRACT Background Predicting activities of daily living (ADL) in stroke patients optimizes discharge planning, which relies on accurate functional assessment. Recent studies have shown that functional connectivity (FC) of brain networks induced by upper extremity rehabilitation robotic training (UE-RAT) effectively reflects functional status, but its prognostic value for ADL remains unclear. Objective Utilize functional near-infrared spectroscopy (fNIRS) to measure FC during UE-RAT and develop machine learning models to evaluate the predictive value of task-FC for ADL. Methods This study recruited 86 patients with subacute stroke. Activation and FC features of key brain regions, such as the superior frontal cortex (SFC) and primary motor cortex (M1), were measured in the resting state and during UE-RAT using fNIRS. Concurrently, 38 clinical features were collected. With modified Barthel Index (mBI) ≥75 at discharge as the prediction target, machine learning algorithms such as artificial neural network (ANN) were used to construct resting-state fNIRS model, task-state fNIRS model, clinical model, and combined model, and analyze the importance of the predictor variables based on the Shapley additive interpretation (SHAP). Results The combined model constructed by combining clinical and task-state fNIRS features had the best predictive performance (AUC_mean: 0.955, 95% CI: 0.948–0.962). Higher connectivity between the ipsilateral premotor cortex (iPMC) and primary motor cortex (iM1) during the task state, along with higher mBI scores and lower mRS scores, predict significant improvement in functional independence for patients. Conclusions UE-RAT induced FC can be a valid biomarker for mBI prediction and can improve the accuracy of rehabilitation prediction.

Tracing source footprints of heavy metal(oid)s in coastal soils using traditional statistical techniques and machine learning data-driven models.

Moroccan producers of households’ detergents suffered from cash flow bullwhip. This distortion of financial flow originated from the bullwhip effect produced during the COVID 19 pandemic. A recent study confirms that some performance attributes are correlated with the degree of exposure to the CFB. The relative significance of these performance criteria, however, is not immediately apparent. The objective of this research is to develop a multi-criteria mathematical model which will serve as a basis to assess the performance of the companies under study and to define the determinants of Cash Flow Bullwhip control using the MACBETH method. This research is conducted on a sample of Moroccan producers of household detergents. The findings indicate that the importance of financial variables is higher than that of supply-chain elements, and internal control factors. Good supply chain asset management, financial efficiency, financial liquidity, control activities, financial debt, and supply chain credibility are the main levers to control the CFB.

ABSTRACT Improving walking and biking for recreation contributes to enhancing public health. Despite the importance of social and physical variables in promoting walking and biking as recreational activities, few studies have looked at how these domains simultaneously influence these activities. This study aims to investigate the links between walking and cycling as recreational activities and the individual, social, and built environmental factors in the medium-sized southern city of Temuco. The objectives were examined utilising a quantitative method and three types of regression analysis. Despite a finding that people’s attitudes towards cycling are more positive than those towards walking or using a private car, just a small portion of respondents reported using bicycles for recreation. The results show that several factors are related to walking and cycling as recreational activities, such as age, encouragement, companionship, lifestyle, and safety. Urban policymakers may use these findings to promote walking and cycling for recreation.

Accelerating climate change and the intensifying global food security crisis have increased the importance of reliable crop classification across diverse environmental conditions. Existing crop classification models have primarily focused on improving accuracy by learning spectral and temporal patterns from satellite imagery; however, their black-box nature makes it difficult to understand the rationale behind each prediction, limiting their applicability in real-world agricultural decision-making. To address this issue, this study introduces a multimodal Transformer model that incorporates a BERTbased bidirectional attention mechanism, aiming to retain classification performance while enhancing interpretability. The proposed BERT Hybrid model employs a PVT backbone to extract spatial features from Sentinel-2 satellite imagery and integrates them with meteorological time-series embeddings; bidirectional self-attention is then used to jointly model cross-temporal and cross-modal interactions. We further conduct comparative experiments under the same conditions as the MMST-ViT(Multi-Modal Spatial-Temporal Vision Transformer) baseline, evaluating not only overall accuracy but also temporal attention patterns across crop growth stages and the relative importance of different weather variables. Experimental results show that bidirectional attention alleviates excessive focus on specific timestamps or single variables, producing more consistent and interpretable attention distributions. This study highlights the performance– interpretability trade-off in multimodal agricultural AI models and provides a foundation for building trustworthy deeplearning systems for crop monitoring. In addition, because the proposed approach relies solely on globally accessible Sentinel-2 satellite imagery and publicly available meteorological data, it demonstrates the potential for constructing large-scale crop monitoring systems at low cost, aligning with the principles of appropriate technology.

The intelligent transformation in the manufacturing industry poses challenges to traditional quality control methods, particularly in handling redundant data and ensuring model interpretability within high-dimensional, multivariate assembly processes. This study presents an integrated approach combining Principal Component Analysis (PCA), Back Propagation neural network (BP neural network), and permutation importance to improve quality prediction and traceability in the automotive body-in-white rear panel dimensional chain. The data for this study originates from the actual production process of an automotive manufacturer. It comprises direct geometric measurements from the rear panel of a specific vehicle model’s Body-in-White (BIW). The measurement points from key coordinates that influence rear panel matching serve as the numerical input variables. The corresponding measurement result from the Skeleton Assembly is utilised as the output variable, which represents the final assembly quality and is treated as a numerical variable in this model. PCA is first applied to reduce dimensionality and eliminate data redundancy. Then, two types of neural networks—single and sequential—are constructed to model nonlinear relationships, with the single neural network exhibiting superior performance in accuracy (average R2 > 95%) and generalisability (RMSE < 0.1). To address the lack of interpretability in conventional neural networks, the permutation importance of variables is assessed to pinpoint the primary sources of bias and to clarify the mechanisms of variable interactions. The automotive company’s practical validation demonstrates the model’s capability to predictively assess the effects of abrupt alterations in bodyside dimensions on rear panel matching quality. The close agreement between predicted (e.g., 1.053693) and actual (e.g., 1.01) values confirms model accuracy, diminishing the reliance on supplementary quality control resources. This study provides a traceable, data-driven framework for enhancing quality control in complex manufacturing assemblies.

To explore, identify, and develop novel blood-based indicators using machine learning algorithms for accurate preoperative assessment and effective prediction of postoperative complication risks in patients with rheumatoid arthritis (RA) undergoing total knee arthroplasty (TKA). A retrospective cohort study was conducted including RA patients who underwent unilateral TKA between January 2019 and December 2024. Inpatient and 30-day postoperative outpatient follow-up data were collected. Six machine learning algorithms, including decision tree, random forest, logistic regression, support vector machine, extreme gradient boosting, and light gradient boosting machine, were used to construct predictive models. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), F1-score, accuracy, precision, and recall. SHapley Additive exPlanations (SHAP) values were employed to interpret and rank the importance of individual variables. According to the inclusion criteria, a total of 1 548 patients were enrolled. Ultimately, 18 preoperative indicators were identified as effective predictive features, and 8 postoperative complications were defined as prediction labels for inclusion in the study. Within 30 days after surgery, 453 patients (29.2%) developed one or more complications. Considering overall accuracy, precision, recall, and F1-score, the random forest model [AUC=0.930, 95% CI (0.910, 0.950)] and the extreme gradient boosting model [AUC=0.909, 95% CI (0.880, 0.938)] demonstrated the best predictive performance. SHAP analysis revealed that anti-cyclic citrullinated peptide antibody, C-reactive protein, rheumatoid factor, interleukin-6, body mass index, age, and smoking status made significant contributions to the overall prediction of postoperative complications. Machine learning-based models enable accurate prediction of postoperative complication risks among RA patients undergoing TKA. Inflammatory and immune-related blood biomarkers, such as anti-cyclic citrullinated peptide antibody, C-reactive protein, and rheumatoid factor, interleukin-6, play key predictive roles, highlighting their potential value in perioperative risk stratification and individualized management.

Abstract This paper presents a novel approach to variable selection in small area estimation, focusing on the Fay–Herriot model. Traditional methods, such as those based on Akaike and Kullback symmetric divergence information criteria, often rely on stepwise selection and focus on the complete model, without individually examining the influence of auxiliary variables. The Shapley value of cooperative game theory is proposed to measure the average importance of auxiliary variables. The Shapley value evaluates all possible combinations of auxiliary variables, ensuring an efficient average influence of the selected combination. We study its properties mathematically and investigate its performance through simulation experiments, showing consistent identification of the true generating variables even under challenging conditions. An application to the 2022 Spanish Living Conditions Survey illustrates the method’s usefulness in selecting the model on which to base predictors of poverty proportions in Spanish provinces by sex.

“GenAI” assistants are gradually being integrated into daily tasks and learning, but their uptake is no less contingent on perceptions of credibility or safety than on their capabilities per se. The current study hypothesizes and tests its proposed two-road construct consisting of two interface-level constructs, namely perceived transparency (PT) and perceived safety/guardrails (PSG), influencing “behavioral intention” (BI) both directly and indirectly, via the two socio-cognitive mediators trust in automation (TR) and psychological reactance (RE). Furthermore, we also provide formulations for the evaluative lenses, namely perceived usefulness (PU) and “perceived risk” (PR). Employing survey data with a sample of 365 responses and partial least squares structural equation modeling (PLS-SEM) with bootstrap techniques in SMART-PLS 4, we discovered that PT is the most influential factor in BI, supported by TR, with some contributions from PSG/PU, but none from PR/RE. Mediation testing revealed significant partial mediations, with PT only exhibiting indirect-only mediated relationships via TR, while the other variables are nonsignificant via reactance-driven paths. To uncover non-linearity and non-compensation, a Stage 2 multilayer perceptron was implemented, confirming the SEM ranking, complimented by an importance of variables and sensitivity analysis. In practical terms, the study’s findings support the primacy of explanatory clarity and the importance of clear rules that are rigorously obligatory, with usefulness subordinated to credibility once the latter is achieved. The integration of SEM and ANN improves explanation and prediction, providing valuable insights for policy, managerial, or educational decision-makers about the implementation of GenAI.

Coastal landscapes are complex socio-ecological systems that undergo rapid transformations driven by both natural dynamics and human pressures. Their sustainable management depends on robust, cost-effective remote sensing methodologies for long-term monitoring and quantitative assessment of spatiotemporal change. In this study, we developed an integrated remote-sensing-based framework that combines historical aerial photograph interpretation, transition matrix analysis, and machine learning to assess coastal dune landscape dynamics over a seventy-year period. Georeferenced orthorectified and preprocessed aerial imagery freely available from the Italian Ministry of the Environment for the years 1954, 1986, and Google Satellite Images for 2022 were used to generate detailed land-cover maps, enabling the analysis of two temporal intervals (1954–1986 and 1986–2022). Transition matrices quantified land-cover conversions and identified sixteen dynamic processes, while a Random Forest (RF) classifier, optimized through parameter tuning and cross-validation, modeled and compared landscape dynamics within protected Long-Term Ecological Research (LTER) sites and adjacent unprotected areas. Model performance was evaluated using Balanced Accuracy (BA) to ensure robustness and to interpret the relative importance of change-driving variables. The RF model achieved high accuracy in distinguishing change processes inside and outside LTER sites, effectively capturing subtle yet ecologically relevant transitions. Results reveal non-random, contrasting landscape trajectories between management regimes: protected sites tend toward naturalization, whereas unprotected sites exhibit persistent urban influence. Overall, this research demonstrates the potential of integrating multi-temporal remote sensing, spatial statistics, and machine learning as a scalable and transferable framework for long-term coastal landscape monitoring and conservation planning.

Accurate prediction of shrimp body weight is critical for optimizing harvest timing, feed management, and stocking density decisions in intensive aquaculture. While prior studies emphasize environmental factors, operational management variables—particularly harvesting metrics—remain understudied. This study quantified the predictive importance of harvesting-related variables using 5 years of industrial-scale operational data from 12 ponds (5479 cleaned records, 34.94% retention rate). We trained seven machine learning models and applied three independent feature importance methods: consensus importance ranking, SHAP explainability analysis, and Pearson correlations. Main findings: Operational variables (days of culture: 2.833 SHAP, stocking density: 1.871, cumulative feed: 1.510) ranked substantially above environmental variables (temperature: 0.123, pH: 0.065, dissolved oxygen: 0.077). Partial harvest frequency showed bimodal clustering, indicating two distinct viable operational strategies. The Weighted Ensemble model achieved the highest performance (R2 = 0.829, RMSE = 4.23 g, MAE = 3.12 g). Model stability analysis via 10-fold GroupKFold cross-validation showed that the Artificial Neural Network (ANN) exhibited the tightest confidence bounds (0.708 g width, 27.7% coefficient of variation), indicating exceptional consistency. This is the first study to systematically analyze the importance of harvesting variables using SHAP explainability, revealing that operational management decisions may yield greater returns than marginal environmental control investments. Our findings suggest that operational optimization may be more impactful than environmental fine-tuning in well-managed systems.

Chronic abdominal pain is prevalent among individuals with inflammatory bowel disease (IBD). Substantial progress has been made in the past few decades in understanding how pain signals travel from the gut to the brain. Preclinical studies have greatly contributed to uncovering the mechanisms of chronic pain in IBD. However, practical applications in clinical settings are still lacking. In this Review, we present the current understanding of the pathological mechanisms of chronic pain in IBD. We explore the neuroplastic changes that occur along the afferent pain pathway from the nerve endings in the colonic mucosa to the brain. We explore the importance of new variables in chronic IBD pain, including the influence of host-microbe interactions during disease, and the effect of comorbid psychological factors, including stress and anxiety, in triggering pain. Finally, we discuss gaps in treatment strategies while proposing new directions for research to identify novel priorities that could pave the way for effective pain therapies.

Cardiovascular diseases are a leading cause of global mortality, with hypertension, obesity, and other factors contributing significantly to risk. Artificial Intelligence has emerged as a valuable tool for early detection, offering predictive models that outperform traditional methods. This study analyzed a dataset of 709 individuals from Ecuador, including demographic and clinical variables, to estimate cardiovascular risk. During preprocessing, records with missing values and duplicates were removed, and highly correlated variables were excluded to reduce multicollinearity and prevent overfitting. The performance of several machine learning algorithms-including Decision Trees, Random Forest, Gradient Boosting, Extreme Gradient Boosting, LightGBM, Extra Trees, AdaBoost, and Bagging-was compared, while addressing class imbalance using SMOTE and a hybrid ROS-SMOTE approach. Gradient Boosting with the hybrid technique achieved the best performance, obtaining an accuracy of 0.87, a precision of 0.81, a recall of 0.74, and an F1-score of 0.75. Its superior performance is attributed to its sequential error correction mechanism and integrated regularization strategies, which effectively reduce overfitting and improve generalization in noisy or imbalanced datasets. These findings demonstrate the potential of AI-based models to improve early detection and management of cardiovascular disease, highlighting the importance of anthropometric, clinical, and blood pressure variables in predicting cardiovascular risk.

ABSTRACT Aim As anthropogenic activities continue, species are exposed to climate change and rapid urbanisation that alter their distribution; however, the relative contributions of climate and anthropogenic influence on species distribution are unknown for most species. We use an environmentally sensitive salamander genus ( Eurycea ) that occupies urban and forested habitats to test the relative importance of temperature‐, precipitation‐ and urbanisation‐related variables before placing these results into a phylogenetic context. We aim to test the impact of climate and urbanisation in driving the distribution of the genus while evaluating patterns of niche conservatism. Location Nearctic. Taxon Eurycea (Plethodontidae, Lungless Salamanders). Methods We developed MAXENT niche models for 13 Eurycea species using bioclimatic and urbanisation‐related variables. We assessed the importance of these environmental variables through permutation importance and compared response curves to determine niche overlap. Phylogenetic analyses tested for evolutionary constraints on species responses to environmental factors. Results Climatic variables were the primary drivers of Eurycea distributions, while urbanisation‐related variables had lower overall importance. Phylogenetic analyses revealed that responses to urbanisation‐related factors, specifically impervious surface and human population density, exhibited significant phylogenetic signal, indicating a stronger evolutionary constraint on responses to urbanisation than to climate. While climatic variables showed limited phylogenetic conservatism, niche overlap analyses demonstrated that more closely related species had greater similarity in ecological responses to urbanisation than to climate. Evolutionary history influenced species' ecological tolerances, with some environmental responses more conserved than expected under Brownian motion. Main Conclusions Our findings highlight the role of evolutionary history in shaping Eurycea responses to environmental variation. While climatic factors predominantly influence broadscale distributions, urbanisation‐related responses are more evolutionarily conserved across the genus. These results suggest that past evolutionary trajectories may constrain species' capacity to adapt to novel anthropogenic stressors, underscoring the importance of incorporating phylogenetic perspectives in conservation strategies for Eurycea and other evolutionarily constrained taxa.

Purpose While controversies regarding the use of midpoint in Likert scale measurements pervade in survey research, the existing literature mainly focuses on confirmatory analyses aiming to assess effects on scale reliability and factorial structure. The consequences of midpoint removal in predictive contexts – which serve as the primary foundation for practical interventions – remain underexplored. We investigate this concern by delving into the stability of the statistical properties of three prediction models built upon the Theory of Planned Behavior framework, by simulating scenarios with and without a midpoint in survey data originally measured on a 7-point Likert scale. Design/methodology/approach We conduct 5,000 partial-least-squares regression models to simulate the effects of the replacement of the midpoint responses in the original datasets based on several replacement rules with psychological plausibility. We compare the results of the original models with the results of the 5,000 simulations. Findings Reduced scales lead to a lower degree of measurement reliability and a lower explanatory power of the models. The statistical significance of predictors is affected under some replacement scenarios. The estimated coefficients also change, to the point that the importance of the independent variables in the original model differs from the one obtained in simulated data. Also, the statistical significance of the predictors, as recommended by the Theory of Planned Behavior, is unstable, and the estimated coefficients of the 5,000 simulation models tend to either overestimate or underestimate the original coefficients. Research limitations/implications We raise awareness that although building survey research on a strong and stable theoretical background is necessary, it may not be sufficient in practical contexts. If the research design is not properly calibrated in terms of measurement, unreliable results may serve as grounds for costly, but likely inefficient, practical interventions. Originality/value Unlike previous research focused on testing the effects of the midpoint on the reliability of a scale and its factorial structure, all driven by confirmatory analysis aiming to test the conformity of the data with a specific theory, we explore potential effects of the midpoint removal in predictive settings, often used in practical interventions.

Progress in the study of the emission characteristics of intermediate and semivolatile organic compounds from motor vehicles.