A k-Nearest Neighbor Parallel Searching Approach Using Various-Width Clustering

Abstract

Over recent decades, database sizes grown due to
the large used in many industrial application. Due to different
size and type of information present today, it creates the interrupt
to find the exact result of respective query. A k-nearest neighbor
searching techniques used to find the nearest result of the query
point. In this proposed system, a parallel k-nearest neighbor
approach, based on various-width clustering, to find the nearest
neighbor with parallel approach. In various-width clustering, it
creates the different size radius clusters from high, medium and
low-dimensional data sets, arrange them in set of distance and
cluster ID. The Proposed parallel k-NN search technique based
on GPU using threads, create efficient result in minimum time
space. This approach, minimizes the comparison time required
to find distance within the clusters, in addition, it also reduced
the computational cost and increase the speed up for searching