Comparison of Kernel Functions in Support Vector Machine (SVM) Method for Phishing Website Detection

Muhammad Farhan Said (Halu Oleo University)
Norma Muhtar (Universitas Halu Oleo)
Bahriddin Abapihi (Halu Oleo University)
Arman Arman (Halu Oleo University)
Muhammad Kabil Djafar (Halu Oleo University)
Herdi Budiman (Halu Oleo University)

Journal of Science and Technology
Vol 6, No 1: Maret 2026

Phishing is a rapidly growing form of cybercrime that poses a significant threat to internet users. Therefore, an effective and accurate detection method is essential. This study aims to analyze and compare the performance of several kernel functions within the Support Vector Machine (SVM) method for detecting phishing websites. The kernel functions used include linear, polynomial, Radial Basis Function (RBF), and sigmoid. The dataset is evaluated using stratified 5-fold cross-validation, with F1-score as the primary performance metric. The results show that the RBF kernel achieves the highest performance, with the best F1-score of 94.64%, followed by the polynomial kernel. The linear kernel performs reasonably well on nearly linearly separable data, while the sigmoid kernel shows the lowest and least stable performance. Based on these findings, the RBF kernel is recommended as the most suitable choice for phishing website detection using the SVM method.