OPTIMIZATION OF DENSITY BASED CLUSTERING DBSCAN USING BAT ALGORITHM

Abstract

Clustering Algorithms are used for the task of classifying spatial databases and also in many other applications like data mining etc. It groups the points such that points within a single group have similar characteristics. There are many clustering algorithms are available for different type of applications. One of them is density based clustering DBSCAN that is used for identifying arbitrary shape of clusters based on their density along with noisy outliers. Secondly, recently many Bio inspired algorithms are used for solving the optimization problems and many other real world complex problems. Bat algorithm is one of the bio-inspired techniques used for solving optimization problems in various fields. It is basically inspired by the echolocation behavior of bats especially micro bats. Bat adjusts its frequency and wavelength accordingly to find its prey’s position. In this proposed work, hybrid of bat algorithm and DBSCAN is used to improve the cluster quality and also time complexity. For achieving this, first the best position of bats in search space is found out i.e. cluster center points, further it groups the other points using those cluster centers i.e. making the clusters according to their density using DBSCAN approach. Results of this work are improved intra cluster distance of clusters and also reduced time complexity of DBSCAN. It may take some extra time to calculate the best position i.e. cluster centers.