Extensive Analysis of Clustering Algorithm for Large Datasets Using Density-Based Clustering and Swarm Intelligence

Abstract

Clustering is the approach of unsupervised learning. The clustering process efficiently handles large scale dataset and maintains the purity of clusters—the density-based clustering algorithm can manage large scale dataset. The approach of clustering faces problems correctness and computational overhead.This paper improves the DBSCAN algorithm with partial probability function, increases cluster correctness, and reduces computational overhead. The applied swarm intelligence algorithm on a density-based algorithm manages the different clusters of clusters to merge with the same centres and EPS point. The process of particle swarm intelligence also reduces the noise and boundary value of data points and increases core points' value. This paper also studies stream data clustering. The stream data clustering modified the MCM algorithm with threshold function, the modified MCM Handel the evolved feature concept. These all algorithms were implemented in MATLAB software and applied high-dimension datasets such as a flame spiral pathway. The results' analysis indicates that the modified algorithm improves the correctness of cluster data approx. 2-4%.