Variable neighborhood search for automatic density-based clustering
Fatima Boudane, Ali Berrichi · 2017
Clustering is a well studied data mining task. Many effective clustering algorithms have been proposed over the years. However, most of them suffer from the problems of arbitrary shaped data. Although, density-based clustering algorithms can identify clusters of arbitrary shapes and handle noise (or outliers) very effectively, one difficulty is that it is hard to choose parameter values, such as the density threshold. In this paper, we propose a variable neighborhood search heuristic to handle arbitrary shaped data automatically, without using any parameter values. This heuristic has been tested on seven selected datasets from the literature. It appears that it produces high quality clustering. Using a clustering validation index as fitness criteria, experiments show that our approach achieves better performance compared with six other clustering methods from the literature.