Search Results (23,984)

Femicide is one of the most extreme manifestations of gender-based violence and remains a persistent challenge across Latin America. This study examines the association between socioeconomic and contextual factors and femicide rates across 12 Latin American countries during 2014–2024 using an unbalanced panel dataset. The dependent variable is the femicide rate per 100,000 women, while the explanatory variables include the Gini index, GDP per capita, female labor force participation, and intentional homicide rates. Owing to heteroskedasticity, serial correlation, and contemporaneous dependence across panels, the model was estimated using Panel-Corrected Standard Errors (PCSE) with a common AR(1) disturbance structure. The Gini index was negatively and statistically significantly associated with femicide rates (β = −0.057, p = 0.004), whereas intentional homicide rates showed a positive and statistically significant association (β = 0.059, p < 0.001). GDP per capita retained a negative but non-significant coefficient (β = −0.371, p = 0.076), while female labor force participation was also negative and non-significant (β = −0.008, p = 0.556). These findings indicate that the associations between femicide and socioeconomic conditions are sensitive to model specification, while income inequality and generalized lethal violence remain statistically significant correlates. The results underscore the multidimensional nature of femicide and the need for cautious interpretation of aggregate cross-national associations. Full article

To encourage the use of hydrogen energy and improve the dependability of hydrogen fuel cells, this study develops a two-stage semi-supervised framework for anomaly pattern discovery and codification. Isolation Forest, One-Class Support Vector Machine, and Local Outlier Factor are first applied to unlabeled time series data; six supervised classifiers are subsequently trained on ensemble-derived pseudo-labels to learn these patterns efficiently. The analysis uses a reproducible 20% sample (random seed 42) of 185,721 observations from four fuel cells in the NASA Prognostics Data Repository, yielding 37,144 observations. The three unsupervised algorithms exhibit distinct detection patterns, with 280 common detections (15.07% of each model’s flagged observations and 0.754% of the sample). A consensus-weighted ensemble identifies 1296 potential anomalies (3.49%). Random Forest best reproduces the ensemble pseudo-labels, with 99.365% testing accuracy; this value measures pattern learnability and is not independent verification of physical faults. Gini importance identifies power difference, capacity, load power, load current, measured power, and load resistance as the principal predictive variables. Because independently confirmed fault labels and maintenance outcomes are unavailable, the reported detections are interpreted as systematic deviations requiring electrochemical or expert verification. Full article

This study investigates how lithological origin influences the engineering performance, stripping resistance, and multi-scale material characteristics of aggregates used in asphalt pavement applications. Three representative aggregate systems—carbonate limestone from Cebeci (Istanbul), siliciclastic sandstone from Arnavutköy (Istanbul), and mafic volcanic basalt from the Karatepe Formation (Çorlu, Tekirdağ)—were comparatively evaluated. The experimental program included physical and mechanical tests, X-ray diffraction, inductively coupled plasma optical emission spectroscopy, inductively coupled plasma mass spectrometry, thin-section petrography, and stripping resistance testing. Physical, mechanical, and durability test results were evaluated using replicate measurements and are reported as mean ± standard deviation where applicable. Basalt showed the best mechanical and durability performance, with the lowest flakiness index, Los Angeles abrasion value, and magnesium sulfate soundness loss, but exhibited the lowest unmodified retained coating. Limestone showed a higher inferred binder-affinity potential, interpreted from its calcite-dominated mineralogy, whereas sandstone displayed higher sensitivity related to particle shape, mineralogical heterogeneity, phyllosilicate-bearing components, and petrographic pore/microvoid characteristics. The addition of 0.5% CL 90 S hydrated lime produced an intermediate improvement in stripping resistance, increasing retained coating values to 75–80% for all aggregates, whereas 1.0% hydrated lime further improved retained coating to 80–85%. Higher lime contents of 1.5% and 2.0% did not provide additional improvement under the applied laboratory conditions. These results indicate that 1.0% CL 90 S hydrated lime represents the most effective dosage among the tested conditions. Overall, the results demonstrate that aggregate performance is governed by coupled lithological controls involving mineralogy, petrography, geochemistry, inferred surface-related interaction potential, mechanical durability, stripping resistance, and anti-stripping additive response. The proposed framework should therefore be interpreted as a comparative multi-scale assessment rather than a direct predictive model. Full article

Photocatalytic wastewater treatment is a promising technology for degrading persistent organic pollutants; however, its optimization remains challenging because photocatalytic performance depends on complex interactions among catalyst properties, operating conditions, and wastewater composition. Machine learning (ML) has emerged as a powerful tool for accelerating catalyst development, predicting photocatalytic performance, and optimizing process parameters. This review critically analyzes recent studies on ML-assisted photocatalytic wastewater treatment, covering supervised learning, ensemble learning, deep learning, and hybrid optimization approaches for predicting degradation efficiency, reaction kinetics, and catalyst performance. Rather than simply summarizing existing studies, the review compares the strengths, limitations, and applicability of different ML models while evaluating the influence of dataset quality, feature engineering, and validation strategies on predictive reliability. Emerging developments in explainable artificial intelligence, physics-informed machine learning, digital twins, and autonomous catalyst discovery are also discussed. Current challenges, including limited datasets, data heterogeneity, model overfitting, lack of standardized benchmarking, and poor transferability to real wastewater systems, are critically examined. Finally, future perspectives emphasizing standardized datasets, interpretable AI, rigorous model validation, and intelligent catalyst design are proposed. This review provides a practical roadmap for integrating artificial intelligence with photocatalysis to accelerate the development of reliable and sustainable wastewater treatment technologies. Full article

Environmental interpretation (EI) in protective forests remains an underexplored field, despite its demonstrated potential for fostering emotional place attachment and conservation attitudes. This study aims to design and evaluate in situ the “Magical Portal of Biodiversity and Culture” self-guided interpretive trail at the “Los Arrayanes” Protective Forest (16 ha, Montúfar Canton, Carchi Province, Ecuador), a native arrayan forest (Myrcianthes hallii) deeply rooted in the ancestral territory of the Pasto-Tuza people, and to formulate educommunicational strategies. The trail design is grounded in Tilden’s interpretive principles, Morales’s topic–theme structure, and Ham’s TORE model, and it formulates Environmental Education and Communication Strategies. A qualitative, multimethod design was employed, with in situ field observation, participatory action research (PAR) workshops with the de Monteverde community, and in-depth interviews with seven key informants (March–April 2026); data were analyzed through the hermeneutic interpretive method and validated through categories, method, and theory triangulation, member checking, and researcher reflexivity. The analysis yielded five emergent categories: (1) vital and heritage place attachment; (2) psychological restoration and sensory reconnection; (3) environmental curiosity and empathy; (4) situated conservation learning; and (5) community governance and intergenerational education. Additionally, it identified six educommunicational strategies: (1) sensory learning; (2) storytelling; (3) participatory interaction; (4) guided self-learning; (5) situated learning; and (6) the biocultural approach. These findings provide empirical evidence that biocultural interpretation promotes conservation attitudes and community participation. In practical terms, this research offers protected-area managers, community organizations, and interpretation planners a model replicable for designing self-guided interpretive trails in biodiverse territories where living biocultural heritage is at risk. Full article

Climate change and urbanization are intensifying urban flood risks worldwide, making flood risk management critically important. While machine learning has been widely applied for flood susceptibility mapping, many studies do not integrate socioeconomic dimensions for comprehensive risk assessment; moreover, traditional subjective evaluation methods provide insufficient quantification of expert judgment uncertainty. This study presents a hybrid framework for urban flood risk mapping, which integrates interpretable machine learning and the Z-number-based Fuzzy Analytic Hierarchy Process (Z-FAHP), applied to Tokyo, Japan, utilizing a high-resolution digital elevation model (DEM) derived from airborne LiDAR and other publicly available geospatial datasets. The framework leverages machine learning efficiency while better accommodating vague linguistic expert judgments by incorporating confidence levels. Flood susceptibility was derived through machine learning with 14 features, with model performance rigorously evaluated using both non-spatial and spatial 4-fold cross-validation, and the CatBoost model demonstrated optimal performance. SHAP (SHapley Additive exPlanations) analysis was further employed to enhance model transparency by quantifying feature contributions. Exposure and vulnerability were quantified from 3 and 5 socioeconomic indicators, respectively, through Z-FAHP. These three criteria were synthesized to produce the flood risk map using simple additive weighting. Results reveal that approximately 32.9% of the study area faces high-to-very-high flood risk, concentrated in eastern lowlands and river corridors. This transferable framework can be applied to other cities globally as an effective tool for urban flood risk management. Full article

Continental shale reservoirs exhibit pronounced multi-scale heterogeneity, with reservoir quality governed by the interplay of lamina assemblages, bedding continuity, and lithological spatial variability. Conventional digital outcrop modeling (DOM) primarily targets geometric reconstruction and three-dimensional (3D) visualization of sedimentary bodies, which is insufficient for the fine-scale structural characterization and quantitative modeling required for shale reservoirs. Here we present a Digital Shale Outcrop Modeling method (DSOM) tailored to continental shale reservoirs, exemplified by the Gulong Shale in the Qingshankou Formation (Upper Cretaceous) of the Songliao Basin, northeastern China. DSOM integrates six sequential modules: digital outcrop reconstruction, digital section interpretation, virtual well construction, virtual well correlation, 3D structural modeling, and parameter extraction. A high-precision digital outcrop model covering 0.369 km² was constructed from 1885 calibrated UAV images (DJI Mavic 3 Enterprise) using Structure-from-Motion (SfM) photogrammetry, yielding derived products including a digital outcrop model (DOM), digital surface model (DSM), digital elevation model (DEM), orthomosaic, and dense point cloud. Seven virtual wells were extracted along the outcrop strike, and a unified lithological classification comprising seven lithotypes was established. A regionally persistent rusty-yellow ferruginous siltstone layer served as a marker bed for virtual well correlation. Results reveal a distinct vertical lithological transition: the section above the marker bed is dominated by muddy deposits, with black mudstone and dark-gray silty mudstone collectively accounting for 52.84% of the total area, whereas the section below the marker bed exhibits a marked increase in silt-grade material, with gray siltstone reaching 32.78%. This vertical evolution reflects a depositional shift from relatively high-energy to low-energy conditions. Using the virtual wells as conditioning data, 3D lithological probability models for all seven lithotypes were constructed via Sequential Indicator Simulation (SIS), achieving quantitative representation of lithological spatial distribution and lateral variability under bedding constraints. Our results demonstrate that DSOM effectively converts outcrop digital information into reservoir structural data, providing reliable constraints for multi-scale structural characterization, 3D geological modeling, and heterogeneity evaluation of the Gulong Shale. More broadly, DSOM establishes a methodological framework for digital characterization of fine-grained sedimentary reservoirs, bridging the gap between outcrop-scale observations and subsurface reservoir modeling. Full article

Building facades are continually exposed to weathering, material aging, hygrothermal cycling, and service-related disturbances. Cracks, delamination, spalling, and seepage can compromise durability and safety, while associated thermal anomalies provide indirect evidence of deterioration. Deep learning, UAV inspection, infrared thermography, LiDAR, building information modeling (BIM), and digital twins have shifted facade inspection toward automated sensing and spatially referenced assessment. Yet most studies remain focused on isolated measures of model accuracy and give limited attention to the full pathway from detection and quantification to component localization, condition rating, and repair. This review organizes the evidence along five dimensions: annotation granularity (A0–A2/A2*), visual task hierarchy (T1–T5), fusion level (F0–F3), spatial mapping level (S0–S3), and engineering maturity (M1–M3). The synthesis indicates that acquisition conditions, material heterogeneity, negative-sample composition, and annotation rules constrain model generalization. Single-modality methods cover classification, detection, segmentation, and partial geometric quantification, but facade-specific external validation remains limited. Multimodal gains depend on registration quality, thermophysical conditions, defect mechanisms, and sensor reliability. Mapping two-dimensional outputs to point clouds, BIM, and digital twins is technically feasible, but error propagation, component matching, and condition-rating protocols remain inconsistent. Interpretability tools and vision foundation models may support verification, annotation, and reporting but cannot replace dedicated systems with validated error bounds. Future work should prioritize cross-dataset benchmarks, facade-specific lightweight models, reproducible multimodal evaluation, standardized 2D-to-3D accuracy protocols, uncertainty-aware manual review, and engineering condition-rating frameworks. Full article

Transport mode choice in intermediate Latin American cities remains insufficiently documented, particularly where mobility surveys contain limited information on alternative-specific transport attributes. This study examined the associations between the reported transport mode choice and the available traveler, trip, temporal, and origin–destination characteristics in Loja, Ecuador. The analysis used 1642 records of respondent-selected daily trips from an origin–destination survey. Because respondents reported up to three transport modes in the chronological order of use, the principal mode was reconstructed using a capacity-based hierarchy rather than assuming that the first reported stage represented the main mode. Multimodality was defined as the presence of at least two distinct valid modes. The hierarchy classified 49.21% of records as public bus, 41.72% as private motorized transport, 6.09% as active and micromobility, and 2.98% as commercial or other motorized transport. The modal category changed for 12.12% of records relative to the first-stage definition, while 37.82% of records were multimodal. The principal multinomial model used a reduced three-category hierarchy-defined outcome and achieved an accuracy of 0.776, macro-F1 of 0.631, and macro balanced accuracy of 0.736. Vehicle ownership showed the largest adjusted association with individual or other motorized transport relative to collective transport (RRR = 16.10; 95% CI: 11.60–22.40), followed by driving license possession (RRR = 3.69; 95% CI: 2.61–5.22). Across 20 repeated Random Forest partitions, the mean accuracy was 0.814, although the recall for active and micromobility remained low at 0.138. The results demonstrate that the outcome definition materially affects the modal shares, coefficient estimates, and predictive performance. Because the trip distance and comparable alternative-specific level-of-service attributes were unavailable, the findings should be interpreted as adjusted associations within the analytical sample rather than causal effects or population-level modal shares. Full article

Reliable short-term freeway traffic forecasting is essential for proactive traffic management, including congestion warning, ramp metering support, and traveler information services. However, many existing forecasting models treat spatial dependencies as synchronous or weakly directional, limiting their ability to represent delayed upstream–downstream traffic propagation. This study proposes Traffic-DiMAGNet, a lightweight and interpretable lag-aware directed spatio-temporal graph neural network for real-time freeway flow forecasting. The model constructs a sparse directed sensor dependency graph by integrating physical road adjacency, training-set lead–lag traffic priors, and learnable source–target node embeddings. A lag-aware bidirectional propagation module then shifts inter-sensor messages according to estimated propagation delays, while directed random-walk normalization, directional gating, and multi-scale causal convolutions preserve asymmetric traffic semantics with low computational cost. Experiments on four Caltrans PeMS datasets show that Traffic-DiMAGNet consistently outperforms recurrent, diffusion-based, attention-based, and adaptive-graph baselines regarding MAE, RMSE, and MAPE values. The learned directed lag structures provide interpretable propagation information, and the lightweight architecture supports rolling 5 min forecasting, indicating practical potential for real-time freeway monitoring and proactive traffic management. Full article

University students, as high-quality human capital, play a critical role in promoting the sustainable development of urban systems. Drawing on survey data from 525 university students in Tianjin, China, this study employs eight widely used machine learning algorithms to predict and interpret settlement intentions. The best-performing model was selected based on predictive performance, and the SHapley Additive exPlanations (SHAP) approach was employed to reveal the relative importance and interaction effects of the influencing factors. The results indicate that Random Forest is the optimal predictive model, achieving an 83.8% accuracy rate in predicting settlement intentions. Among the 18 predictors examined, family support, occupation prospects, job opportunities, Hukou attraction, and climate & environment were the five most influential positive factors affecting settlement intentions, whereas hometown distance was identified as the strongest negative predictor. We further analyzed the heterogeneity and interaction effects of these influencing factors. This study provides new insights into systematically explaining university students’ settlement intentions. Full article

Sparse count data, such as patent document–keyword matrices, often contain excessive zeros and overdispersion, making conventional distance-based clustering methods less suitable. This study proposes a zero-inflated likelihood-based graph neural clustering method with a zero-inflated negative binomial likelihood, denoted as ZIL-GNC-ZINB. The proposed method combines graph-smoothed node representations with cluster-specific zero-inflation probabilities and count-intensity parameters. By incorporating graph-neighborhood information into the cluster membership update, ZIL-GNC-ZINB jointly accounts for structural zeros, overdispersion, and local graph relationships among observations. The proposed method was applied to a quantum-computing patent document–term matrix consisting of 9416 patent documents and 82 reduced keywords. Compared with K-means, K-means clustering based on principal component analysis (PCA+K-means), and K-means clustering based on graph convolutional networks (GCN+K-means), ZIL-GNC-ZINB achieved the best performance in terms of negative log-likelihood (NLL), zero area underneath the receiver operating characteristic (ROC) curve (AUC), and zero Brier score. The resulting clusters revealed interpretable quantum-computing sub-technologies, including photonic qubit control, hybrid quantum–classical computing, superconducting qubit hardware, and quantum security networks. Simulation experiments under zero proportions of 0.5, 0.7, and 0.9 further showed that the proposed method becomes increasingly effective as zero inflation becomes more severe. In the extreme zero-inflation setting, ZIL-GNC-ZINB achieved the best performance in NLL, Zero AUC, adjusted rand index (ARI), normalized mutual information (NMI), and clustering accuracy (ACC). These results demonstrate that zero-inflated likelihood modeling combined with graph-based clustering provides an effective and interpretable framework for sparse high-dimensional count data. Full article

Cancer incidence exhibits substantial spatial disparities associated with environmental, behavioral, built-environment, healthcare access, and socioeconomic conditions, yet the extent to which these county-level associations vary geographically and across spatial scales remains insufficiently understood. This study evaluates an explainable spatial epidemiology workflow that integrates Random Forest (RF), SHapley Additive exPlanations (SHAP), permutation importance, Ordinary Least Squares (OLS), Geographically Weighted Regression (GWR), and Multiscale Geographically Weighted Regression (MGWR) to examine county-level incidence for all-site cancer as a composite benchmark, colorectal cancer, female breast cancer, and melanoma of the skin across Texas, USA. All outcomes were screened from a common leakage-safe pool of 44 predictors. RF-SHAP/permutation screening was conducted only in spatially separated discovery counties, and the retained predictors were subsequently evaluated using OLS, GWR, and MGWR in independent confirmation counties. The screening procedure retained 19–23 predictors across outcomes, reducing model dimensionality by approximately 48–57%. Although screening substantially reduced AICc, in-sample also decreased, indicating improved parsimony and complexity-adjusted fit rather than improved explanatory performance; same-cardinality random, correlation-based, and LASSO benchmarks further showed that the advantage of RF-based screening was outcome- and model-dependent. In independent confirmation analyses, GWR was preferred by AICc for all-site cancer (AICc = 442.79), whereas OLS was preferred for colorectal cancer (356.48), female breast cancer (367.50), and melanoma (234.56); MGWR was not AICc-preferred for any outcome. However, where estimation was feasible, MGWR provided complementary multiscale information by distinguishing fitted associations characterized by near-global versus more localized spatial bandwidths. Five-fold nested spatial-block cross-validation showed stronger geographic predictive performance for RF, with pooled values of 0.422, 0.235, 0.442, and 0.341 for all-site, colorectal, breast, and melanoma outcomes, respectively, whereas GWR produced negative held-out for all four outcomes. These findings support explainable-ML screening primarily as a transparent dimensionality reduction strategy and demonstrate complementary roles for spatial modeling: AICc evaluates whether additional spatial complexity is justified, MGWR characterizes predictor-specific spatial scales where feasible, and spatial-block validation evaluates geographic predictive generalization. The resulting associations and spatial scales are interpreted as ecological and descriptive rather than causal. Full article

With the ongoing reform of electricity markets and the increasing penetration of demand response (DR) resources, load exhibits stronger price-related uncertainty, posing significant challenges to short-term load forecasting. To address this issue, this paper proposes a DR-oriented forecasting framework that integrates feature decomposition, adaptive response modeling, and deep learning techniques. First, electricity price signals and load data are processed to identify DR-related features and extract representative response characteristics. An adaptive response mechanism (ARM) is then developed to learn time-varying statistical response characteristics from historical load–price states and to construct an explicit price-related correction for the baseline load forecast, while a hybrid CNN-BiLSTM forecasting model is constructed to capture both local spatiotemporal patterns and long-term temporal dependencies. In addition, an optimal weighting strategy is adopted to improve forecasting performance. Case studies demonstrate that the proposed method consistently outperforms conventional forecasting models that neglect DR effects, yielding notable reductions in MAE, RMSE, and MAPE, together with an increase in the coefficient of determination (R²). By effectively quantifying the dynamic price–load interaction, the proposed framework improves forecasting accuracy, robustness, and interpretability, providing a practical solution for short-term load forecasting in electricity markets with DR participation. Full article

Pulmonary hypertension (PH) is a heterogeneous clinical syndrome in which similar haemodynamic abnormalities may arise from distinct vascular, cardiac, pulmonary, thromboembolic, and molecular mechanisms. This complexity limits the ability of conventional classifications and risk scores to fully capture individual disease trajectories and treatment responses. Artificial intelligence (AI) offers a framework for integrating clinical data, electrocardiography, multimodal imaging, invasive haemodynamics, biomarkers, and multi-omics information across the PH care pathway. This review summarises current applications of machine learning and deep learning in early detection, diagnostic referral, right-ventricular and pulmonary vascular phenotyping, molecular endotyping, risk stratification, and therapeutic decision support. Available studies show promising results for AI-assisted electrocardiographic screening, automated echocardiographic and cardiac magnetic resonance analysis, computed tomography (CT)-based phenotyping, and multimodal prognostic modelling. True multi-omics integration in PH remains limited to discovery studies and has not yet yielded externally validated endotype or treatment-response classifiers. Evidence maturity is task-dependent: screening and phenotyping span several PH groups, whereas validated risk tools, molecular endotyping, and pathway-directed therapy remain predominantly PAH-based, particularly in idiopathic/heritable PAH. However, most evidence remains retrospective, derives from selected referral populations, and lacks robust external or prospective validation. No AI-based model currently supports routine drug selection or autonomous clinical decision-making. Future progress will require harmonised multicentre datasets and standardised acquisition protocols, transparent and interpretable models, and prospective studies demonstrating meaningful clinical benefit. AI should therefore be viewed as an emerging decision-support tool that may strengthen precision medicine in PH while complementing clinical expertise across diagnosis, phenotyping, risk assessment, and therapeutic stratification pathways. Full article