TITLE
EVALUATION OF MACHINE LEARNING CLASSIFIERS USING WEKA: FINDINGS FROM MULTI-DOMAIN DATA
AUTHOR(S)
Fatma BÜYÜKSARAÇOĞLU SAKALLI, Özlem AYDIN FİDAN
ABSTRACT
The primary aim of this study is to systematically analyze the performance of classical classification algorithms that are commonly used in machine learning across different datasets and to identify the factors that influence algorithm selection. In the study, Naive Bayes, K-Nearest Neighbor (KNN), Decision Tree (J48), Random Forest (RF), Support Vector Machines (SVM/SMO), and Artificial Neural Networks (ANN) algorithms were implemented using the WEKA software, and four distinct datasets (TURKSTAT Happiness by Gender, Labor, Titanic, and Wine) were examined. The findings revealed that algorithmic performance varies depending on the nature of the dataset. For instance, the Random Forest model achieved the highest accuracy on the Wine Quality dataset, the SMO (SVM) performed best on the Titanic dataset, while Naive Bayes proved to be the most efficient method for the small-scale Labor dataset. Evaluation metrics such as accuracy rate, Kappa statistic, and error measures (MAE, RMSE) enabled a comparative assessment of the models. The main contribution of this study is to present a comprehensive understanding of how classical machine learning algorithms behave across different domain-specific data and to emphasize the importance of data-sensitive algorithm selection. The results are consistent with similar comparative studies in the literature and provide researchers with a methodological framework for model evaluation processes.
DOI: https://www.doi.org/10.70456/VBYB9221
How to cite this article:
Fatma BÜYÜKSARAÇOĞLU SAKALLI*, Özlem AYDIN FİDAN, Sanja Antić, EVALUATION OF MACHINE LEARNING CLASSIFIERS USING WEKA: FINDINGS FROM MULTI-DOMAIN DATA, UNITECH – SELECTED PAPERS - 2025