Soft DBSCAN: Improving DBSCAN clustering method using fuzzy set theory

Clustering is one of the most valuable methods of computational intelligence field, in which sets of related objects are cataloged into clusters. Almost all of the well-known clustering algorithms require input number of clusters which is hard to determine but have a significant influence on the clustering result. Furthermore, the majority is not robust enough towards noisy data. In contrast, density based methods, such as DBSCAN, have obvious advantages over explicit samples. They discover the number of clusters, as well as, they detect noises. Additionally, the shape of such clusters can also be irregular. However, they have difficulties in handling the challenges posed by the collection of natural data which is often vague. This paper presents an efficient clustering technique, named “Soft DBSCAN” that combines DBSCAN and fuzzy set theory. Our new method is galvanized by Fuzzy C Means in the way of using the fuzzy membership functions. The results of our method show that it is efficient not only in handling noises, contrary to Fuzzy C Means, but also, able to assign one data point into more than one cluster. Simulative experiments are carried out on a variety of datasets, throughout different evaluation's criteria, which highlight the soft DBSCAN's effectiveness and cluster validity to check the good quality of clustering results.