Multi-Objective Artificial Bee Colony Algorithm for Parameter-Free Neighborhood-Based Clustering

Introduction

Clustering is one of the most studied data mining processes. Clustering consists to find the intrinsic structure of a dataset, by partitioning the data into distinct and homogeneous groups (or clusters). The objects in the same cluster must be similar to each other and different from objects in other clusters. The most existing clustering algorithms, such as K-Means (Hartigan & Wong, 1979), cannot handle arbitrarily shaped clusters and isolated data. In addition, they depend on the user-defined parameters and suffer from the well-known problem of local minima. Although density-based clustering algorithms (Ester et al., 1996; Zhou et al., 2005; Mishra & Mohanty, 2019; Shi et al., 2018) can find clusters with arbitrary shapes, most of them still cannot handle datasets of varying density and depend on the user-defined parameters.