İbRahim Budak

Aims: Accurate revenue forecasting is essential for effective financial planning in dental healthcare services; however, increasing variability in multidisciplinary clinical workload limits the adequacy of traditional linear approaches. This study aimed to develop and compare explainable machine learning–based models for forecasting monthly revenue in a multidisciplinary public dental clinic and to evaluate the specific contribution of orthodontic production within this framework.Methods: Monthly aggregate data covering 48 months (2021–2024) from dental clinics affiliated with the Kastamonu Provincial Health Directorate were analyzed. Independent variables consisted of monthly procedure volumes in endodontics, restorative treatment, prosthodontics, periodontology, orthodontics, pedodontics, and oral surgery, while total monthly sales constituted the dependent variable. Six regression models were developed using the scikit-learn library: ExtraTreesRegressor, HistGradientBoostingRegressor, ElasticNet, KNeighborsRegressor, MLPRegressor, and a Stacking Regressor combining ExtraTrees and HistGradientBoosting with a Ridge meta-learner. A preprocessing pipeline including RobustScaler and log1p transformation was applied where appropriate. Model performance was evaluated using R², mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE). Model interpretability was assessed using SHapley Additive exPlanations (SHAP).Results: All models achieved acceptable predictive performance; however, the Stacking model provided the best overall fit (R² ≈ 0.95) and the lowest MAE, RMSE, and MAPE (≈18%). Tree-based ensemble methods consistently outperformed distance-based and neural network approaches. ElasticNet demonstrated competitive performance (R² ≈ 0.91), indicating a strong linear component in revenue dynamics. SHAP analysis revealed that prosthodontic and restorative procedure volumes were the dominant drivers of monthly revenue, whereas orthodontic production had a comparatively modest short-term contribution.Conclusion: Ensemble-based machine learning models, particularly stacking approaches, can accurately forecast monthly revenue in multidisciplinary dental clinics using clinical production indicators. Prosthodontic and restorative services represent the primary drivers of short-term revenue fluctuations, while orthodontic activity contributes more modestly within a monthly time frame. These findings support the integration of explainable machine learning models into financial planning and operational decision-making processes in dental healthcare institutions.

This study aimed to evaluate global research trends in scoliosis between 2000 and 2024 using latent Dirichlet allocation (LDA) topic modeling. English-language, peer-reviewed journal articles indexed in Scopus were analyzed using their titles, abstracts and keywords to identify dominant and relatively underrepresented research themes. The findings showed that surgery-related topics accounted for the largest thematic proportion of the scoliosis literature, whereas etiopathogenesis and genetic factors had limited structural representation. Artificial intelligence-related studies increased in recent years but remained primarily focused on radiological assessment, Cobb angle measurement and diagnostic applications. These results demonstrate an imbalance in the thematic distribution of scoliosis research; however, bibliometric frequency and topic proportion should not be interpreted as direct indicators of scientific quality or clinical importance. The findings may help researchers recognize areas that warrant further investigation, particularly etiological and genetic mechanisms. The application of artificial intelligence in genetics, precision medicine and treatment planning should be regarded as a potential direction for future research rather than an outcome directly demonstrated by the present analysis.

Purpose: This study develops and evaluates a multivariate ensemble forecasting framework for tourism revenue that combines Elastic Net and CatBoost under rolling-origin validation, using Türkiye (2002–2024) as an illustrative case to generate medium-term reference scenarios. Design/methodology/approach: Tourism revenue was the dependent variable, with international visitors, Blue Flag beaches, ministry-certified facilities, travel agencies, museums/historical sites, and museum-site visitors as explanatory indicators. The multicollinear multivariate structure was modeled using Elastic Net and CatBoost and evaluated via rolling-origin (expanding-window) cross-validation using RMSE, MAE, and R2. A benchmark ARIMA model was also tested under the same evaluation scheme for comparison. Findings: Elastic Net delivered the best out-of-sample accuracy (R2 = 0.894; RMSE = 4.02; MAE = 2.80), while CatBoost produced a smoother, more conservative trajectory. Standardized Elastic Net coefficients and CatBoost SHAP values consistently identify visitor volume (VIS) and museum visitors (MV) as the dominant drivers. The equal-weight ensemble yields a balanced 2025–2027 nominal reference path with a modest post-pandemic increase. The benchmark ARIMA (2,1,2) performs substantially worse across error metrics, supporting the superiority of the proposed multivariate ML framework. Originality/value: The study contributes a generalizable forecasting workflow that (1) benchmarks regularized linear and gradient-boosted models under a common rolling-origin design and (2) shows that a transparent simple-average ensemble yields more robust medium-term reference scenarios under shocks and structural breaks. Türkiye is used as a demonstration case.

This study aims to map academic trends in sustainable marketing both retrospectively and to develop future projections. To this end, the study presents thematic trend forecasts extending to 2030 by creating a quantitative map of the sustainable marketing literature around five main themes. Only English articles, conference papers, book chapters, and related document types were selected using the query "sustainable" AND "marketing" in the Elsevier Scopus database; abstracts of 17,747 records covering the period 1980–2025 were analysed. After preprocessing, BERTopic was applied using paraphrase-multilingual-MiniLM-L12-v2 embeddings. In line with prior BERTopic practice, the outlier/noise cluster (Topic = −1), accounting for 4.4% of documents, was excluded, and the remaining 209 topics were consolidated into five higher-level themes based on content similarity. The resulting themes cluster around axes such as green marketing, sustainable branding, marine and food-focused sustainability, green supply chain marketing, and AI-supported sustainability applications, quantitatively revealing the current backbone structure of the sustainable marketing literature. Annual publication counts were calculated for each theme and converted into time series, which were then forecasted for the 2026–2030 period using LSTM and GRU models. During the test period, LSTM demonstrated clearly superior performance compared to GRU, with lower errors (MAE and RMSE) and higher R² values. The findings clearly show that themes such as sustainable branding, green supply chain marketing, and AI-supported sustainability applications will experience a strong upward trend in the coming years. In contrast, areas such as sustainable packaging marketing will follow a relatively balanced, mature course. By combining topic modelling with deep learning-based time series forecasting, the study offers a unique methodological contribution to sustainable marketing research. It produces concrete, foresight-based insights for both the academic agenda and brand strategies.

In higher education, academic performance metrics have become a fundamental resource in universities’ strategic planning and quality assessments. The rankings that reveal the success levels of universities are generally determined by parameters based on international standards. However, the methods used in these rankings need to be analyzed in more detail and the current performance needs to be improved strategically. The aim of this study is to classify the academic performance of universities established after 2000 in Türkiye using the Louvain clustering method based on University Ranking by Academic Performance (URAP) indicators. In the study, universities were divided into six different clusters (C1-C6) by analyzing data for the years 2022-2024. These clusters were determined according to indicators such as academic productivity, research capacity and international cooperation. The Louvain method allows universities to be classified according to their similar characteristics and clearly reveals differences in performance. The analysis shows that universities in the Marmara and Aegean regions are concentrated at the top, while those in the Eastern and Southeastern Anatolia regions are concentrated at the bottom. The study provides a strategic framework for improving the current performance of universities and emphasizes the need to support low-performing universities and encourage international cooperation.

Background: Graf’s introduction of hip ultrasonography (US) in the 1980s coincided with a major change in developmental hip dysplasia (DDH) screening and diagnostic practice, with a gradual reduction in reliance on examination- and radiograph-only strategies. Methods: A natural language processing pipeline and non-negative matrix factorization (NLP–NMF) were applied to Scopus-indexed DDH publications, extracting topics and weights across 4 epochs (pre-1980, 1980-1984, 1985-1989, 1980-2024) anchored to the 1980 US milestone. Topic co-occurrence networks, weight distributions, and segmented regression summarized thematic shifts and tested for a 1980 breakpoint. Results: In the post-1980 corpus, a distinct diagnostic imaging theme emerged and strengthened, while surgery declined and anatomy remained influential. In 1985-1989, joint anatomy dominated (39.1%), followed by diagnosis (18.6%), pediatric orthopedics (15.3%), surgery (13.7%), and conservative treatment (13.4%). Across 1980-2024, clinical evaluation accounted for 31.5% of publications, followed by anatomy (19.6%), genetic anomaly (16.7%), diagnostic imaging (16.6%), and surgery (15.6%). Within diagnostic imaging, US carried a high weight (0.948), underscoring its central role. Segmented regression identified a structural breakpoint around 1980, temporally coincident with the period in which hip US entered clinical practice and diffused across the field. Conclusion: The NLP–NMF analysis of 4 decades of DDH literature shows a sustained pivot toward imagingsupported,n clinically anchored evaluation and growing attention to etiologic and syndromic/anomaly contexts alongside anatomical determinants, with potential implications for earlier detection, risk-aligned follow-up, and more targeted use of operative management. Cite this article as: Taşkin R, Yılar S, Budak İ, Uğur F. Ultrasonography and the reorientation of developmental hip dysplasia research: an NLP-NMF trend analysis (1980-2024). Eurasian J Med. 2026, 58 (2), 1249, doi:10.5152/eurasianjmed.2026.251249.

In this study, the prediction performance of different artificial intelligence algorithms was examined using quality of life data from 2016 to 2025. The analysis compared gradient-boosted tree-based XGBoost with LSTM, which has the capacity to model time series and sequential dependencies. In addition, SHAP analysis was applied to ensure the model's explainability and to identify the key factors affecting quality of life. The findings show that both models successfully capture quality of life patterns, with the LSTM model achieving higher out-of-sample accuracy than XGBoost (higher R² and lower MAE, RMSE, and MAPE). SHAP analysis revealed that Purchasing Power and Pollution are the factors with the strongest impact on quality of life. The decisive effect of Purchasing Power indicates that macroeconomic conditions such as real income level, price stability, and Purchasing Power Parity -adjusted welfare indicators directly reflect quality of life. Other factors, such as cost of living, housing price/income ratio, security, healthcare services, climate, and commute time, were found to have varying degrees of importance across countries. These findings emphasize the priority of designing macroeconomic frameworks targeting income/wage policies and price stability alongside policies aimed at improving environmental conditions. The results obtained indicate that policy makers should focus on the efficient allocation of resources. The results obtained provide policymakers with an evidence-based roadmap for the efficient allocation of resources and demonstrate that more detailed analyses can be conducted using different explainable artificial intelligence methods for future research. Additionally, to test the robustness of the model, different training/testing splits, alternative error metrics, and hyperparameter sensitivity analyses were performed; the direction and magnitude of the main findings were found to be consistent across these scenarios. Finally, SHAP-based findings provide a starting framework for policy simulations, enabling the quantitative prediction of potential welfare gains from targeted improvements in specific sub-indices.

Effective medical waste planning relies on the reliable estimation of waste volumes. As operational factors diversify, traditional linear regressions often fail to capture the underlying structure, whereas latent variable–based and ensemble approaches can better represent this complexity. In this study, fine-tuned Partial Least Squares (PLS), scikit-learn–based Gradient Boosting regression (GBR), and a baseline Ordinary Least Squares (OLS) model were compared for estimating medical waste generation using 48 months (2021–2024) of approximate data from Dental Clinics affiliated with the Provincial Health Directorate in Kastamonu. The model inputs were the monthly procedure counts for endodontics, treatment, prosthetics, periodontology, orthodontics, pedodontics, and surgery. Performance was evaluated using Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and R-squared (R²). All models produced accurate predictions; however, PLS provided the strongest fit (R² = 0.979; MAE = 30.488; RMSE = 37.043), outperforming GBR (R² = 0.962; MAE = 36.544; RMSE = 48.990) and the OLS baseline (R² = 0.927; MAE = 41.762; RMSE = 59.013). The findings demonstrate that modern, data-driven waste-management planning is feasible in healthcare institutions and highlight PLS as a robust option, particularly under conditions of small sample size and collinearity.