acsFSDPC: A Density-Based Automatic Clustering Algorithm with an Adaptive Cuckoo Search

Abstract

Clustering has gained increasing attention in the data mining field since it plays an important role in the unsupervised classification of samples. While numerous clustering methods have been proposed, these suffer from various limitations including sensible parameter dependence and difficult identification of the number of clusters. In this paper, an automatic density-based clustering method based on an adaptive cuckoo search (acsFSDPC) was proposed. Data points with higher density and larger distance from other data points are assumed as clustering centers. Firstly, an adaptive cuckoo search algorithm is employed using clustering evaluation index as fitness function to determine the optimal cutoff distance for each cluster number. Then, the cluster number with the minimal fitness function value is chosen as the best number of clusters and the corresponding clustering result is the optimal clustering results. The benefits of acsFSDPC are automatic estimation of cluster number and optimal cutoff distance. Experiments of acsFSDPC and its comparison with other recent methods STClu, ACND, CH-CCFDAC, and LR-CFDP are performed on five simulation data sets and a real data set. Results show that the acsFSDPC is promising in estimating the appropriate number of clusters automatically and effectively.