Improving Cluster Efficiency by Density Based Approach using Hybrid Optics

Clustering is the one of the important methods used in data mining, which has wide range of applications in Image processing, Pattern recognition and Compression. The clustering algorithms are categorized into five methods, such as Partitioning, Hierarchy, Grid-Based, Density Based and Model Based. A new approach based on the Hybrid OPTICS is proposed, which is the Density based method used to cluster the data with high dimension and variable densities. OPTICS is an Ordering Points To Identify the Clustering Structure used to find the density based clusters. It addresses the one of the DBSCAN problem of detecting meaningful clusters in the data of varying densities. OPTICS requires two parameters, ε is used to describes the radius and MinPts is used to describe the number of points require to form the cluster. A point p is a core point if at least MinPts points are found with Epsilon neighbourhood N ε (p). Using this parameters Hybrid OPTICS computes core distance and reach ability distance to find local and global minimum. Hybrid OPTICS automatically determines the border set based on global and local minimum and then performs cluster analysis for each local cluster based on local minimum and integrate the all local minimum to obtain the global minimum. The efficiency of the Hybrid OPTICS is validated using real data set when compared with existing SDE algorithm.