Optimal preference detection based on golden section and genetic algorithm for affinity propagation clustering

Abstract

Affinity Propagation Clustering Algorithm is a well-known effective clustering algorithm that outperforms other traditional and classical clustering algorithms, and the selection of related sensitive parameters (preference, damping factor) is a popular research topic. In this paper, a feasible detecting procedure “GS/GA-AP” based on Golden Section and Genetic Algorithm is proposed to address the aforementioned issue. As a default option, preference is given based on golden section for Affinity Propagation. Unsatisfactory clustering result is robust with selection of preference with Genetic Algorithm. One simulation dataset and five standard benchmark datasets are utilized to verify effectiveness of algorithm we proposed, and the experiment results show that GS/GAAP outperforms traditional Affinity Propagation clustering algorithm.