K-Fold Cross-Validation Analysis for Student Status Prediction Using Classification Models

Predicting the academic status of students in full-scale study processes is crucial for universities in designing institutional planning and accurate resource allocation. It can also help identify students who may require early intervention, thereby avoiding serious strategic impacts of student status on institutional sustainability. Machine learning (ML), a branch of artificial intelligence, can be used to predict the academic status of students by considering the factors that continuously influence their academic performance. This research aims to develop a model that offers optimal student status predictions for universities. To achieve this goal, researcher developed an approach for predicting student status using four models: Artificial Neural Network (ANN), Random Forest (RF), Decision Tree (Tree), and Naive Bayes (NB). In order to optimize the processing of the dataset, researcher used K-fold Cross Validation (KCV) for handling imbalanced datasets. This approach yielded the best results when the original imbalanced dataset was transformed using K-fold Cross Validation (KCV) technique. The RF method demonstrated a dominant advantage over other methods, achieving the highest accuracy score of 92% on the 4th fold. However, the other methods also showed significant performance improvements, with accuracy scores ranging from high to low, including 91.9% (Decision Tree), 89.6% (Naive Bayes), and 84.8% (Artificial Neural Network).