Haydar Koç

Obesity is a growing public health concern, particularly among university students who are exposed to lifestyle changes, disordered eating habits, and reduced physical activity. The aim of this study is to classify obesity risk levels among university students using machine learning classification methods and to identify the most influential factors associated with this risk. The study sample consisted of data collected from 445 students studying at Çankırı Karatekin University. In this context, eight machine learning algorithms—Logistic Regression, Random Forest, Extra Trees, Support Vector Machines, K-Nearest Neighbor, Quadratic Discriminant Analysis, Naive Bayes, and Multilayer Perceptron—were compared to classify obesity risk. Class imbalance in the dataset was addressed using the Adaptive Synthetic Sampling (ADASYN) method applied exclusively to the training set. The models were evaluated using standard performance metrics, and the highest accuracy rate (96.26%) was achieved by the Random Forest model, followed by Logistic Regression with 94.77% accuracy. Variable importance analysis indicated that age, internet use scale score, and fast-food consumption frequency were the most influential factors in classification, while the low correlation between variables (|r| < 0.2) suggested that model performance was driven by the combined contribution of multiple features. Overall, the findings demonstrate that the balanced machine learning approach, particularly ensemble-based methods, can classify obesity risk with high accuracy and provide valuable insights for targeted prevention strategies among university students.

Real-world systems frequently exhibit hierarchical multipartite graph structures, yet existing graph neural network (GNN) approaches lack systematic frameworks for hyperparameter optimization in heterogeneous multi-level architectures, limiting their practical applicability. This study proposes a Bayesian optimization framework specifically designed for heterogeneous GNNs operating on three-level graph structures, addressing the computational challenges of configuring partition-aware architecture. Four GNN architectures—Graph Convolutional Networks (GCNs), Graph Attention Networks (GATs), Graph Isomorphism Networks (GINs), and GraphSAGE—were systematically evaluated using Gaussian Process-based Bayesian hyperparameter optimization with inter-partition message-passing mechanisms. The framework was validated on the TIMSS 2023 dataset (10,000 students, 789 schools, 25 countries), demonstrating that Bayesian-optimized GraphSAGE achieved the highest explained variance (R² = 0.6187, RMSE = 71.73, MAE = 64.32) compared to seven baseline methods. Bayesian optimization substantially improved model performance, revealing that two-layer architectures optimally capture cross-partition dependencies in three-level structures. GNNExplainer was used to identify the most influential student-level features learned by the model, providing explanatory insight into how the model represents individual characteristics. The optimization framework is broadly applicable to other heterogeneous and multi-level graph settings; however, the empirical findings, such as the optimal architecture depth, are specific to hierarchical graphs with structural properties like the TIMSS topology.

Zeynep Yegin, Gokhan Sarisoy, Cumhur Avsar, Ayse Erguner Aral, Haydar Koc. Clin Psychopharmacol Neurosci 2025;23:278-85. https://doi.org/10.9758/cpn.24.1242

Objective: The human genome contains numerous active transposable elements (TEs), including the inserts of young Alus such as AluYa5 and AluYb8. The rapid, human lineage-specific evolution of brain size may be influenced by natural selection and the genomic variation of TEs, which can act as catalysts of evolution. Given their potential effects on human brain development, Alu elements may also be relevant in exploring complementary approaches for psychiatric disorders. This study analyzed the potential impacts of three specific Alu insertions (AluYa5-ACE, AluYa5-PROGINS, and AluYb8-MUTYH) on the susceptibility to panic disorder (PD). Materials and Methods: Alu insertions were determined with Polymerase Chain Reaction (PCR) method in 120 PD patients and 120 healthy control individuals and analyzed with logistic regression. The associations of the investigated Alus with both categorical and quantitative demographic-clinical variables including the Panic Disorder Severity Scale (PDSS) were analyzed using Fisher's exact and Kruskal-Wallis tests. Results: AluYa5-PROGINS and AluYb8-MUTYH did not show an association with panic disorder risk. For AluYa5-ACE, a statistically significant association was determined; logistic regression analysis proved the highly protective effect (47.9%) of I/D genotype against PD compared with D/D genotype (p=0.030, OR=0.521, 95% CI=0.290-0.937). Alu insertions were not associated with panic disorder severity (p>0.05). Conclusion: This is the first combined study in which the potential effects of three Alu insertions in terms of both PD susceptibility and possible association with several disease parameters were evaluated. Further studies about the roles of the Alu elements in the etiology PD are highly recommended.

Zeynep Yegin, Gokhan Sarisoy, Ahmet Uzun, Haydar Koc. Clin Psychopharmacol Neurosci 2024;22:655-61. https://doi.org/10.9758/cpn.24.1201

The negative binomial regression model is a widely adopted approach when dealing with dependent variables that consist of non-negative integers or counts. This model serves as an alternative regression technique for addressing issues related to overdispersion in count data. Typically, the maximum likelihood estimator is employed to estimate the parameters of the negative binomial regression model. However, the maximum likelihood estimator can be highly sensitive to multicollinearity, leading to unreliable results. To eliminate the adverse effects of multicollinearity in the negative binomial regression model, we propose the use of a jackknife version of the Kibria–Lukman estimator. In this study, we conducted a theoretical comparison between the proposed jackknife Kibria–Lukman negative binomial regression estimator and several existing estimators documented in the literature. To assess the performance of the proposed estimator, we conducted two simulation studies and performed a real data application. The results from both the simulation studies and the real data application consistently demonstrated that the proposed jackknife Kibria–Lukman negative binomial regression estimator outperforms other estimators.

Quantile regression is one of the alternative regression techniques used when the assumptions of classical regression analysis are not met, and it estimates the values of the study variable in various quantiles of the distribution. This study proposes ratio-type estimators of a population mean using the information on quantile regression for stratified random sampling. The proposed ratio-type estimators are investigated with the help of the mean square error equations. Efficiency comparisons between the proposed estimators and classical estimators are presented in certain conditions. Under these obtained conditions, it is seen that the proposed estimators outperform the classical estimators. In addition, the theoretical results are supported by a real data application.

Purpose: Sport is an important tool in improving the quality of life of individuals with disabilities. The study aim to examine the life quality of individuals with physical disabilities with different functionality levels in relation to sports. Methods: A total of 36 individuals with physical disabilities participated in the study. The participants included individuals without any regular exercise (non-sportive, NS), those who played Boccia (B), and persons who played Wheelchair basketball (WCB) who could not provide independent ambulation. The Rivermead Mobility Index (RMI), Functional Independency Measurement (FIM), Trunk Impairment Scale (TIS), and World Health Organization Quality of Life Instrument for People with Intellectual and Physical Disabilities (WHOQOL-DIS) were used for evaluating motor impairment of the trunk, functionality levels, mobility, and quality of life (QoL), respectively. Results: Group B had lower RMI scores compared to group NS (p0.05), but higher in the WCB group (p

Poisson regression is used to deal with count data. The Poisson regression ratio estimator of the population mean is extended from single to double sampling. This is made possible by the provision of the population mean of an auxiliary variable. The mean square errors of the proposed estimators are expressed up to the first order. Theoretical and numerical results demonstrate that the proposed double-sampling Poisson-regression ratio estimator has a lower mean square error than the double-ratio and the single-sampling estimator. For Covid-19, the minimum mean square errors yielded by the proposed estimator of the total number of cases are 0.095 cases per day and 67.8 cases, compared with 0.112 cases per day and 84.8 cases with the double-ratio estimator.

Poisson regression model is the most common method used to model count response in many studies. This paper proposes a new ratio-type estimators based on Poisson regression in simple random sampling. The mean square error (MSE) equation of these estimators is obtained in this study. Theoretically, the MSE of the proposed estimators and the MSE of the traditional ratio estimators are compared. As a result of these comparisons, it has shown that the suggested estimator is more efficient than the traditional estimators. In addition, the results of the application part supported the theoretical findings of the study.