Data Partitioning Technique to Enhance DBSCAN Clustering Algorithm

Abstract

Among density- based clustering techniques ,DBSCAN is a typical one because it can detect clusters with widely different shapes and sizes, but it fails to find clusters with different densities and for that we propose a new technique to enhance the performance of DBSCAN on data with different densities ,the new solution contains two novel tech¬niques ,one is the separation (partitioning ) technique that separate data into sparse and dense regions, and the other is the sampling technique that produce data with only one density distribution. the experimental results on synthetic data show that the new tech¬nique has a clustering