Performance Comparison of K-nearest Neighbor, Decision Tree, and Random Forest Methods for Classification of Cyber Defense Master Scholarship Recipients

Abstract

Cyber defense education is essential for developing a workforce capable of addressing evolving cyber threats, particularly in the military sector, where interconnected systems are vital for secure communication and command. This research aims to enhance the selection process for the Cyber Defense Master Scholarship at the Republic of Indonesia Defense University by employing machine learning algorithms. The study compares the performance of K-Nearest Neighbor (KNN), Decision Tree, and Random Forest for classifying eligible scholarship candidates. The results reveal a clear performance hierarchy: KNN achieves a moderate accuracy of 80.68%, offering simplicity and interpretability but lacking the precision of other models. The decision Tree performs with high accuracy (98.86%) but shows vulnerability to overfitting, which may impact generalizability to unseen data. Random Forest emerges as the most robust model, achieving the highest precision and overall stability, with minimal compromise on other metrics. Given the scholarship’s selection requirements, Random Forest is recommended for tasks needing high accuracy and resilience against overfitting, while KNN and Decision Tree offer suitable alternatives for simpler, more interpretable applications.