3 September 2025 in Kahramanmaraş Sütçü İmam Üniversitesi Mühendislik Bilimleri Dergisi
Nicotine is known not only for its addictive properties but also for its potential to alter the genetic structure of cancer cells. This study investigates the genetic alterations caused by chronic nicotine exposure in HCC38 breast cancer cells through a multi-layered computational analysis. Using K-means clustering and seven supervised machine learning algorithms, differentially expressed genes were grouped into three clusters and used as class labels in classification models. Logistic Regression achieved the highest performance with 98.76% accuracy and an F1 score of 0.9869. The Friedman test was applied to evaluate the statistical significance of performance differences among classifiers, and multi-class ROC curves were used to demonstrate their discriminative power. The findings indicate that nicotine exposure leads to genetic reprogramming and activates inflammation-related gene pathways in specific cellular subpopulations. These results highlight the utility of machine learning methods in uncovering biologically meaningful gene expression patterns in cancer research.
23 February 2026 in Turkish Journal of Mathematics and Computer Science
Profiling educators based on their attitudes toward technology and digital competence provides valuable guidance for developing targeted professional development strategies. This study grouped mathematics teacher educators according to their levels of perceived usefulness, perceived ease of use, intention to use technology, and technological proficiency. An open-access dataset was analyzed using K-Means, K-Means++, and Hierarchical Clustering algorithms, resulting in three distinct participant profiles. K-Means was specifically employed to enhance the stability and convergence of initial centroids in the clustering process. These profiles were then used as target labels in supervised classification tasks using five machine learning algorithms: Support Vector Machine (SVM), Random Forest, Decision Tree, K-Nearest Neighbors (KNN), and Naive Bayes. Among these algorithms, the SVM model achieved the highest accuracy of 92%. To assess the performance of the classification models, additional evaluation metrics such as precision, recall, F1-score, AUC, and the Friedman test were employed. The Friedman test showed that SVM consistently outperformed other models, confirming its superior classification capability. The findings underline the value of data-driven approaches in educational technology research and contribute to the optimization of teacher education programs by providing insights into teacher profiling based on technology acceptance and digital competence.