Clustering of noisy regions (CNR)—A GIS anchored technique for clustering on raster map

Abstract

In this proposed work, a GIS anchored system has been approached, which initially takes as input, a digitized map, generally of a very large region, with the population of the common mass in different wards/areas fed as associated data and finally it suggests the most suitable locations for constructing a number of utility service centers. This target can be achieved by formation of clusters of the wards, where only the adjacent regions are the part of any particular cluster. Particularly, for the third world countries like India, not only the heavily populated wards, but also some small populations such as small slum areas, generally locating outskirts of the major cities, which could be viewed as noise due to their small populations, would also have to be considered for the purpose. One such popular and well-accepted clustering technique handling noise is DBSCAN Ester et al. (A density-based algorithm for discovering clusters in large spatial databases with noise, 1996). But, the major demerit of DBSCAN algorithm is that it cannot take as input, the number of clusters to be generated. This is quiet impractical, as because the number of centers to be constructed is decided beforehand, by some planning committee. Moreover, DBSCAN simply omits the noise areas. But, with a motto to provide equal opportunity for every citizen, care should be taken for all parts of the population. The proposed technique takes a step forward towards the solution.