반딧불알고리즘을 적용한 군집 지능 클러스터링 방법

초록

Swarm Intelligence (SI) is a kind of artificial intelligence (AI) discipline that is popular in these days. Fireflyalgorithm (FA) is recently developed for new SI. The objective of this paper is to propose the FA to find theoptimal data clustering solution with number of clusters. Our proposed FA is suitable for highly nonlinear andmultimodal optimization because FA can consider the attractiveness and distance between the fireflies (solutions)simultaneously. We use the silhouette as valid index to evaluate the solution with number of clustersconsidering the intra-cluster and inter-cluster distance at the same time. We also use the relative distance rate( ) of each data j which is used to generate the good initial solutions to save the much computation time forsilhouette evaluations. The performance of our proposed FA using silhouette with relative distance rate isvalidated using several real data sets by experiment and analysis.

키워드