Homogeneous Densities Clustering Algorithm
Ahmed Fahim · International Journal of Information Technology and Computer Science · 2018
Clustering based-density is very attractive research area in data clustering.DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm is the pioneer in this area.It can handle varied shaped and sized clusters.Also, it copes with noise and outliers efficiently, however it fails to handle clusters with varied densities due to the global parameter Eps.In this paper, we propose a method overcomes this problem, this method does not allow large variation in density within a cluster and use only two input parameters that will be called minpts and maxpts.They govern the minimum and maximum density of core objects within a cluster.The maxpts parameter will be used to control the value of Eps (neighborhood radius) in original DBSCAN.By allowing Eps to be varied from one cluster to another based on density of region this make DBSCAN able to handle varied density clusters and discover homogeneous clusters.The experimental results reflect the efficiency of the proposed method despite its simplicity.