An evolutionary clustering method for arbitrary shaped data sets

Clustering is an important tool for data analysis in both scientific and real-world applications. However, most of the existing clustering methods still face two challenges, such as clustering arbitrary shaped data sets and automatically detecting the number of clusters. This paper aims to solve the two challenges. An evolutionary arbitrary shape clustering (EASC) method is proposed for this purpose. In EASC, the path distance is used to measure the similarity between data points and a modified Modularity index is utilized as the optimization objective. EASC is applied to six benchmark problems and compared with some state-of-the-art clustering methods. The experimental results suggest that our approach not only successfully detects the correct cluster numbers but also achieves better accuracy for most of the problems.