Search Results (4,549)

This study evaluates the operational and performance outcomes of railway freight market liberalization across 28 European countries from 2013 to 2024, moving beyond simple operator counts to analyze key traffic, economic, and infrastructure metrics. To control for structural variation, we classified countries into small, medium, and large networks using K-means clustering based on railway route length. A Mixed-Design Repeated-Measures ANOVA was conducted to analyze temporal trends, network-scale impacts, and their interactions. The results reveal that while the total number of active operators rose steadily after 2018, other performance indicators showed an inverse relationship from 2020 onward. The network cluster had a highly significant main effect across all performance indicators, with the largest effect sizes for infrastructure access charges (partial η² = 0.638), freight train-kilometres (partial η² = 0.610), and net tonne-kilometres (partial η² = 0.585). The interaction between year and network cluster was also statistically significant, most notably for net tonne-kilometres (partial η² = 0.583). Underperformance was prominent in small networks, where market entry did not prevent long-term declines in traffic volume, whereas large networks achieved growth of over 25% in net tonne-kilometres. These quantitative findings show that counting active operators alone is an inadequate measure of liberalization success, as market performance and overall activity levels are strongly associated with network scale and capacity rather than market fragmentation. Full article

Architectural colour in public libraries significantly influences indoor atmospheric quality and occupant emotional well-being, yet palette selection in design practice remains predominantly reliant on subjective intuition, precedents, and vendor catalogues. This study develops a computational framework that links multimodal image colour extraction, circular clustering, human perceptual evaluation, and multi-criteria decision weighting into a structured analysis pipeline. Using a corpus of 500 curated library interior images across diverse geographic regions, a multimodal large language model extracted dominant architectural colour records and visual weights, which were deterministically converted to HSV space and clustered using weighted circular k-means. A semantic differential experiment involving fifty design students across five bipolar scales provided empirical perceptual ratings, evaluated through repeated nested ten-fold cross-validation across five machine learning algorithms. The analysis identified five recurrent colour paradigms, demonstrating that spatial hue distributions form distinct chromatic clusters across modern library architecture. Perceptual models revealed that average colour features reliably predict perceived warmth and quietness, whereas modernity and naturalness depend more heavily on non-chromatic spatial cues. By integrating these predictive models with analytic hierarchy process and Delphi expert weights, the framework establishes a transparent decision-support protocol for early schematic design. This allows architects and clients to quantitatively compare alternative colour schemes against functional zoning requirements, providing an objective, accountable foundation for evidence-based interior design. Full article

In social media contexts, recommendation systems are significant for enhancing the delivery of personalized content. More traditional recommendation systems aim primarily for recommendation accuracy, but do little on the topic related to the explainability and transparency of recommendation decisions. This paper presents an explainable user-based collaborative recommendation framework by leveraging K-Means clustering, similarity-based collaborative filtering, neighbor voting, and generative Large Language Model (LLM) explanations. The framework firstly utilizes K-Means clustering to cluster users with relatively similar behavioral traits. Pearson Similarity and Cosine Similarity are then utilized to select the Top-N similar users for each cluster. To generate personal recommendations for the content themes and content types, we employed similarity-based neighbor voting. Finally, Gemini 2.5 Flash is included as an explainability layer to create brief human-readable explanations of recommendation outputs. The findings presented a high behavioral similarity among adjacent users, which displayed Mean Pearson Similarity = 0.8044, Mean Cosine Similarity = 0.8113, and Mean Combined Similarity = 0.8078. It was found that a content-type recommendation performed better than a preferred-theme recommendation in social media platforms with large amounts of content. In general, this framework enhances interpretability, transparency, and the user understanding of recommendation while still providing efficient collaborative recommendation performance. It identified behaviorally similar users, and generated human-readable explanation suggestions through Gemini 2.5 Flash. Full article

The increase in the number of people living after an oncological diagnosis requires the use of instruments capable of assessing both general dimensions and specific aspects of quality of life. The study aimed to evaluate the psychometric properties of the Romanian version of the Quality of Life in Adult Cancer Survivors (QLACS) questionnaire, analyze the relationships between its scores, age, and residence environment, and identify distinct quality of life profiles. Methods: A cross-sectional study was conducted in 224 adult women with a confirmed diagnosis of breast cancer. The internal consistency of the 12 QLACS domains was assessed by Cronbach’s alpha coefficient. The structure of the instrument was explored by principal component analysis with Oblimin rotation. Differences according to residence were analyzed by the Mann–Whitney test, and the relationships with age by the Spearman coefficient. Multivariable regression models including age and residential environment were fitted for all 12 QLACS domains. Quality of life profiles were identified by hierarchical classification and k-means analysis. Results: Cronbach’s alpha coefficients ranged from 0.832 to 0.925, indicating good internal consistency across the 12 QLACS domains. Principal component analysis retained 11 components, explaining 73.738% of the total variance, providing preliminary exploratory support for the multidimensional structure of the instrument. Most domains were clearly delineated, whereas Cognitive problems partially overlapped with Recurrence distress. Urban residence was associated with higher Cognitive problems scores (r = 0.17), whereas rural residence was associated with higher Family distress scores (r = 0.29). Age was negatively correlated with Negative feelings (rho =−0.200, p = 0.003). A weak exploratory association was also observed with Recurrence distress (rho = 0.135, p = 0.043), although the corresponding multivariable model was not statistically significant. Cluster analysis identified three exploratory quality-of-life profiles, and residential environment was associated with cluster membership, although the association was small (Cramer’s V = 0.188). Conclusions: The Romanian QLACS showed good internal consistency and preliminary exploratory structural evidence in women with breast cancer. Further studies are needed to confirm its psychometric properties. Full article

While ecological restoration revitalizes degraded urban lands, biophysical recovery does not inherently guarantee public visual acceptance. This study introduces a survey-free computational framework to evaluate the visual acceptability of restored landscapes, analyzing 685 post-restoration photographs from South Korea’s Ecosystem Conservation Levy Return Projects. We integrated SegFormer and MiDaS to extract 22 depth-weighted, eye-level landscape metrics and categorized sites into three distinct visual-composition typologies via K-means clustering. To approximate aesthetic response, a CLIP vision–language model generated CLIP-derived, image-level preference scores. A Random Forest model (in-sample R² = 0.882) and SHAP analysis were then applied to decode the non-linear relationships between physical landscape elements and the CLIP-derived scores. The results revealed an asymmetric association structure: visual evidence of degradation (bare soil) and urban intrusion (buildings) were strongly associated with lower estimated preference, whereas care-signaling elements (unpaved trails) were associated with a markedly positive effect. SHAP dependence analysis identified favorable conditions for higher estimated preference, including bare soil exposure below 10%, minimal building visibility, and a baseline visual diversity (SHDI) above 1.0–1.1—a positive association that was sustained up to the dataset’s maximum observed complexity (~2.0). By translating visual landscape composition into quantitative, interpretable thresholds, this survey-free framework offers a scalable and explainable foundation for integrating aesthetic considerations into evidence-based ecological restoration design. Full article

Urban heat islands pose increasing risks to human settlements, yet the differential mechanisms by which grey–green spaces regulate diurnal and nocturnal land surface temperature across local climate zones remain insufficiently understood. This study addresses this gap through a Hangzhou case study, integrating a ten-indicator grey–green space system with two local spatial identification approaches—K-means clustering and an LCZ-inspired simplified scheme—and a Random Forest-SHAP framework. The LCZ-inspired scheme outperformed K-means clustering, with a mean diurnal–nocturnal Test R² of 0.4344 across twelve models, compared to 0.2977 for K-means. Diurnal and nocturnal LST were driven by systematically different factors: building density dominated daytime LST in most LCZ types (22.0% to 27.8%), while canopy height dominated nighttime LST (22.7% to 30.2%), revealing a systematic shift from building-dominated daytime to vegetation-dominated nighttime. This shift did not occur in compact built-up areas, suggesting that built-up density may be a threshold condition for the shift. Key variables exhibited nonlinear threshold effects with saturation points varying by LCZ type: canopy height cooling saturated at approximately 4 m in LCZ2 but required 17–21 m in LCZ3 and LCZA. These SHAP-based patterns and turning points should be regarded as exploratory, sample-dependent associations evaluated within the training data; their spatial stability across held-out regions was not assessed. Factor interactions were interval-dependent rather than globally fixed. Spatial cross-validation confirmed that random splitting substantially overestimated model performance, highlighting the necessity of spatially explicit validation. The methodological framework provides a replicable approach for urban thermal environment research and offers LCZ-specific threshold hypotheses for thermal regulation planning in subtropical megacities, subject to further spatial and cross-city validation. Full article

The development of teachers’ digital competence has become a key issue in contemporary education systems, particularly in contexts where technology integration remains uneven. This study aims to profile teachers’ digital competence through a cluster analysis based on pedagogical digital competence, digital integration, and contextual conditions, and to examine its relationship with selected professional and demographic variables. A quantitative, non-experimental, cross-sectional design was employed with a sample of 299 teachers from Educational District 16-02 in the Dominican Republic. Data were collected using a questionnaire based on the DigCompEdu framework. The internal consistency of the instrument was assessed, and principal component analysis (PCA) with Varimax rotation was applied to the nine composite dimension scores for dimensionality reduction, yielding a three-component solution. Subsequently, cluster analysis (hierarchical and k-means) was conducted to identify differentiated teacher profiles. The results indicate the existence of three distinct profiles characterized by different levels of pedagogical digital competence, integration of technology in teaching, and contextual conditions. Statistically significant associations or differences were identified for educational level, teacher age, and student age, whereas no statistically significant associations or differences were identified for gender or duration of technology use in the classroom. Within the studied educational context, these findings highlight the heterogeneity of teachers’ digital competence profiles and suggest the potential value of considering both pedagogical and contextual factors when designing targeted professional development strategies. The study is limited to selected professional and demographic variables, which should be interpreted as contextual correlates rather than broad sociodemographic determinants. Full article

The Northeastern Donkey represents an important animal genetic resource adapted to the semi-arid conditions of Brazil, being recognized for its hardiness and ability to survive under challenging environmental conditions. However, the progressive reduction in the use of these animals in productive activities has contributed to population decline and increased concerns regarding the loss of genetic diversity. In this context, the present study aimed to characterize the morphometric variability of a Northeastern Donkey population from the Brazilian semi-arid region. A total of 51 adult donkeys were evaluated using linear and circumference measurements obtained with a measuring tape and a hippometer. Data were analyzed using principal component analysis (PCA), hierarchical and non-hierarchical cluster analyses (k-means), analysis of variance, and canonical discriminant analysis. A descriptive three-component PCA solution summarized body circumference and overall body size, limb robustness, and animal stature and linear body dimensions, explaining 74.49% of the total variance; however, parallel analysis supported only two components, so the three-axis interpretation remains exploratory. Cluster analysis identified three morphostructural profiles corresponding to small-, intermediate-, and large-sized animals. Classification consistency was subsequently assessed using canonical discriminant analysis, which achieved 98.0% apparent classification accuracy and 88.2% accuracy under leave-one-out cross-validation. The small and large profiles occurred exclusively in Santa Quitéria and Aquiraz, respectively, indicating potential facility-related influences. The results demonstrate phenotypic variability within the evaluated sample and contribute to a better understanding of the morphological diversity of the Northeastern Donkey. Furthermore, the identification of distinct morphostructural profiles provides valuable information for conservation initiatives, population characterization, and the sustainable management of this threatened local genetic resource. Full article

Urban–rural transition zones (URTZs) represent critical spatial units where urban expansion interacts with ecosystems. In arid regions, however, the response of ecological environmental quality (EEQ) to rapid urban expansion and spatial restructuring remains insufficiently understood. Taking Xinjiang as a representative arid-region case, this study develops an analytical framework integrating dynamic URTZs identification, EEQ assessment, and the analysis of influencing factors and nonlinear responses to systematically investigate URTZs expansion and EEQ changes across 13 typical urban agglomerations from 2002 to 2022. URTZs were identified using K-means clustering by integrating population density, nighttime light intensity, and impervious surface information. An improved remote sensing ecological index (ARSEI) was then developed by incorporating the abundance index (AI) into the traditional RSEI framework. Finally, XGBoost and SHAP were employed to identify the key determinants of EEQ and reveal their nonlinear responses and interactions. The results showed that: (1) URTZs expanded rapidly and continuously from 2002 to 2022, with their total area increasing by more than threefold and exhibiting a spatial restructuring pattern characterized by expansion from central cities toward multiple nodes. (2) Despite the rapid expansion of URTZs, overall EEQ remained at a relatively high level; however, the grade structure exhibited a trend of “expansion at both ends and contraction in the middle,” intensifying the spatial differentiation of EEQ. (3) XGBoost and SHAP analyses identified precipitation (PRE), population density (POP), digital elevation model (DEM), and slope as major factors explaining the spatial variation in EEQ. Interaction analysis further revealed strong interactions between PRE × DEM and PRE × POP. High EEQ values were primarily distributed in areas characterized by favorable precipitation conditions, moderate elevations, and gentle terrain, indicating that the synergistic effects of hydrothermal conditions and topographic constraints play a significant role in shaping EEQ in URTZs. These findings demonstrate that rapid URTZs expansion in arid regions does not necessarily lead to an overall decline in EEQ but may intensify its spatial differentiation. Therefore, ecological governance of URTZs should shift from a singular focus on controlling urban expansion toward differentiated spatial management that jointly considers hydrothermal conditions, topographic constraints, population concentration, and ecological carrying capacity, thereby promoting coordinated urbanization and ecological conservation. Full article

Artificial intelligence (AI) recommender systems operationalize personalized marketing by transforming behavioral data into individualized choice architectures, yet greater model complexity or personalization granularity need not improve every service objective. This study compares population-level item-neighborhood recommendation, behaviorally segmented neighborhood recommendation, and neural collaborative filtering (NCF) across 776,741 deduplicated 1–5 ratings from 646,576 anonymized users and 541 Coursera courses. The primary warm-start evaluation is restricted to 10,296 users (1.59% of the user population) with sufficient interaction history; the remaining sparse-history users contribute to fitting but not to the within-user confirmatory estimand. Configurations are selected only on three validation seeds and then frozen before guarded final evaluation across ten deterministic user-aware seeds. No rank-capable model dominates across objectives: KNN provides the strongest RMSE and personalized-neighborhood coverage profile, ClusteredKNN achieves the strongest observed Top-K retrieval with lower neighborhood support, and a metadata-rich NCF variant achieves the lowest MAE at substantially greater fitting cost. The rating-only UserMean baseline further shows that low numerical prediction error need not imply useful item ranking. Overall, the findings support a contingency view of AI personalization: greater segmentation granularity or model complexity does not inherently create greater value; model choice should instead reflect the service objective, available behavioral evidence, coverage tolerance, operating cadence, and governance requirements. Full article

The building sector accounts for 34% of global energy demand, making accurate short-term load forecasting essential. Small commercial buildings remain challenging due to high-frequency noise and data sparsity. This study proposes a hybrid machine learning framework that decouples behavioral load shapes from climate-driven magnitude variations. K-Means clustering identifies operational archetypes, while a dual-stage supervised pipeline predicts daily shapes via a non-linear SVM classifier and peak magnitude via regression. Validated on real-world 15 min office building data from Murcia, Spain, the framework consistently outperforms ARIMAX, Prophet, persistence, and zero-shot foundation models (TimesFM, Chronos-t5). The best configuration achieved R² = 0.83 and MAE = 0.37 kW, confirming that domain-specific behavioral decoupling surpasses general-purpose pretraining in small-scale commercial datasets. Full article

Autistic speech is highly heterogeneous, yet group-level comparisons may obscure meaningful individual acoustic patterns. This study used unsupervised machine learning to identify data-driven acoustic profiles in native speakers of Cypriot Greek, including autistic and neurotypical adults. Participants produced disyllabic pseudowords across controlled phonetic and stress conditions. Sixteen acoustic measures, including fundamental frequency, formants, duration, cepstral peak prominence, Mel-frequency cepstral coefficients, jitter, shimmer, harmonics-to-noise ratio, and intensity, were summarized at the participant level and normalized appropriately. Principal component analysis retained eight components explaining 81.4% of total variance, followed by k-means clustering. A three-cluster solution provided the best silhouette coefficient among tested solutions and showed good bootstrap stability. Cluster membership was significantly associated with diagnostic group: one profile was exclusively autistic, one was relatively balanced, and one was predominantly neurotypical. The dominant acoustic dimension was driven primarily by voice-quality and spectral measures, particularly cepstral peak prominence, intensity, shimmer, harmonics-to-noise ratio, and jitter, whereas pitch and formant measures contributed comparatively little. These findings demonstrate that unsupervised acoustic profiling can reveal stable, diagnostically relevant speech phenotypes that are not captured by conventional binary group comparisons, highlighting substantial within-group heterogeneity in autistic speech and supporting more individualized approaches to characterizing vocal variation. Full article

Increasing anthropogenic pressures and nutrient inputs threaten the ecological status of coastal ecosystems along the Aegean coast of Türkiye. This study evaluated coastal water quality and trophic conditions using physicochemical parameters, nutrient concentrations, chlorophyll-a, national eutrophication criteria, the Trophic Index (TRIX), and multivariate statistical analyses across 25 coastal stations monitored during 2022–2023. While basic physicochemical variables showed limited spatial variability, nutrient concentrations, chlorophyll-a, and TRIX values exhibited pronounced spatial differences among stations (p < 0.05), identifying Bostanlı (Inner İzmir Bay) as the principal eutrophication hotspot. Principal Component Analysis (PCA) revealed two dominant environmental gradients: a Nutrient Enrichment Gradient associated with total phosphorus (TP), ammonium, nitrite, and nitrate nitrogen (32.6% of total variance), and a Hydrographic–Thermal Gradient associated with salinity, conductivity, total dissolved solids, temperature, and dissolved oxygen (30.4%). Hierarchical and k-means clustering analyses further distinguished environmentally coherent coastal sectors and confirmed the clear separation of Bostanlı from the remaining sampling stations. The close agreement among nutrient distributions, chlorophyll-a concentrations, TRIX values, and multivariate analyses indicates that localized anthropogenic nutrient enrichment, rather than basin-scale hydrographic variability, drives trophic differentiation along the coast. These findings demonstrate the value of integrated monitoring approaches combining physicochemical, trophic, and multivariate indicators for identifying eutrophication hotspots and supporting ecosystem-based management in Mediterranean coastal ecosystems. Full article

Reliable prediction and interpretation of photosynthetic rate are important for precision environmental management in greenhouse tomato production, but model stability is often limited by variable redundancy, measurement-session effects, and environmental heterogeneity. This study developed an integrated gas-exchange-data-based framework for key-factor selection, photosynthetic-rate reconstruction, cross-session validation, environmental correction, and physiology-informed limitation screening. Core predictors were first identified from high-dimensional gas-exchange variables using K-means clustering and random-forest importance analysis. A CatBoost–ExtraTrees–RBF-SVR stacked ensemble was then constructed to reconstruct the leaf photosynthetic rate from selected gas-exchange variables, and its cross-session generalization was evaluated using nested cross-validation and leave-one-file-out (LOFO) extrapolation. Environmental correction was further applied to improve cross-session comparability, and rule-based limitation screening was used to classify potential environmental constraints associated with reduced photosynthesis. The stacked model achieved an RMSE of 0.806 and an $R^2$ of 0.9918 under nested cross-validation, and an RMSE of 0.814 and an $R^2$ of 0.9917 under LOFO extrapolation. After environmental correction, the cross-session variability of the environmentally corrected photosynthetic indicator was reduced by 96.0%, reflecting improved cross-session comparability. Under deployable sensor inputs, performance decreased markedly (A1: RMSE = 4.578, $R^2$ = 0.735; A2: RMSE = 4.610, $R^2$ = 0.731), compared with the gas-exchange baseline (RMSE = 0.833, $R^2$ = 0.991). Thus, routine greenhouse sensor models should be regarded as approximate rather than high-accuracy substitutes for the gas-exchange-based model. Full article

Population ageing is reshaping care systems and cultural meanings of later life. Present explorative qualitative study examined how healthcare professionals working with older adults in two Italian cities (Rome and Bergamo), represent old age. Moreover, we explored whether these representations vary by age, gender, and location. Fifty professionals aged 30–60 years, each residing and practicing professionally in either Rome (n = 25) or Bergamo (n = 25), were interviewed using an open-ended prompt on resources, difficulties, and transitions related to old age. A quali-quantitative approach based on Emotional Text Mining (ETM) was applied to verbatim transcripts. The analysis included text preprocessing, bisecting k-means clustering, and correspondence analysis. Seven clusters emerged: Witnessing ageing, Passing years, Challenging old age, Denial, Awareness, Rethinking services, and Old age as a problem. Correspondence analysis identified six factors; the first three, accounting for 60.24% of total inertia, were interpreted as Time (Present vs. Future), Places (Protected Shelter vs. Exposure to Risk), and Care (Assistance vs. Planning). Together, the factors and clusters outlined a symbolic field ranging from direct observation of ageing and reflection on time, to defensive responses and concerns about decline, and to greater awareness of older adults’ needs and calls for more responsive services. No statistically significant differences emerged by age or gender. Conversely, cluster distribution was significantly associated with participants’ location (Rome vs. Bergamo; χ² = 49.84, p < 0.001), suggesting possible differences in the representation of older adults across these two geographical areas. Overall, representations oscillated between proximity and defense, individual responsibility and system-level delegation, supporting reflective training and organizational interventions to strengthen care cultures. Full article