A Modified DBSCAN Clustering Algorithm

摘要

DBSCAN is one of the most famous clustering algorithms that is based on density clustering, it can find clusters of arbitrary shapes. A major limitation of DBSCAN is its sensitivity to the input parameters also it cannot handle data containing clusters of varying densities. In this paper, we introduce some enhancement to DBSCAN algorithm by estimating its parameters based on the number of occurrences of the fifth neighbor distance. Also, we make a merge to the obtained clusters that are close to each other and have same density and have a thick area joining them. Experimental results demonstrate that our algorithm is effective and efficient and outperform DBSCAN in detecting clusters of different densities.