Abstract

Spatial data clustering roles a crucial play in the knowledge discovery of spatial databases. The standstill challenges of clustering methods to the base of amplifying volume and diversity of data are unsuccessful in multivariate shapes, dimension and densities. To improvise this, the proposed multidimensional optimization clustering method named MOC (Optimized K-means with density and distance-based clustering (OKMDDC)) is taken in account to discover the clusters with diverse shapes and densities in spatial databases with contemporaneous multidimensional arbitrary length of data and its attributes. Thence clusters are segregated on facts of dimensional, subspaces and results for spatial applications. This methodology is inclusive of three phases as Extreme Learning Machine (ELM) classifier for dataset reduction of attribute or dimensional, OKMDDC incorporating a K-means algorithm and Improved Cuckoo Search (ICS) to do sub cluster on the facts of small spherical or ball-shaped sub clusters and then accordingly do merge of subcluster with local density. OKMDDC do cluster of base higher density than neighbor subcluster and in accordance with large distance from subclusters with higher densities. Proposed MOC algorithm poses near-linear time complexity in view of the data set size and dimension, and to hit upon cluster with multivariate shapes and densities.

Paper Details