Performance Enhancing Modified DBSCAN Algorithm

Abstract

Density based spatial clustering of application with noise i.e. DBSCAN clustering algorithm is used for large and spatial data and also used in various important applications. In spite of its usefulness drawbacks still exist of the algorithm which is as follows: Two input parameters are used which needs to be stated in prior and exact calculation of these parameters is not feasible. Poor or not exact calculation of input parameters gives less visibility of clusters which results into less efficiency. Clarity in visibility of clusters is one of the main motives while making clusters. Further performance is low in terms of runtime, speed. This paper proposed enhanced algorithm for DBSCAN which resolves all above mentioned issues. This modification is able to increase visibility of clusters and hence reducing noise elements by providing exact calculation of input parameters. In wide variety of data, detection of outliers is very useful in various application domains. By using Gaussian Probability function the run time efficiency of proposed method is better than the existing one and hence increases performance. Algorithm is implemented in Mat lab and experimental results show the effectiveness of proposed technique while applied on various large datasets.