Jurnal Publikasi STMIK Pontianak

Enhancing Classification Performance of k-NN and SVM with Firefly Algorithm

Abstrak

In the realm of machine learning, optimal parameter selection plays a critical role in determining classification performance. Two widely adopted algorithms Support Vector Machine (SVM) and k-Nearest Neighbors (k-NN) are highly sensitive to specific parameters. Manual tuning of these parameters is often time-consuming and may not yield optimal outcomes. This study proposes the use of the Firefly Algorithm (FA), a swarm intelligence method inspired by the flashing behavior of fireflies, to automatically determine optimal parameter values. A set of datasets from the UCI Machine Learning Repository is utilized to evaluate the effectiveness of the proposed approach. For SVM, the parameters tuned include C and gamma, while for k-NN, the optimal number of neighbors (k) is determined. The results demonstrate that FA enhances classification accuracy and produces more stable models due to reduced performance variance. The findings suggest that FA is a viable and efficient solution for parameter tuning in SVM and k-NN, particularly valuable for researchers seeking to construct reliable classification models without the burden of manual configuration.

Kata Kunci: Firefly Algorithm, Parameter Optimization, Support Vector Machine (SVM), k-Nearest Neighbors (k-NN)