Productive K-Nearest Neighbor (PKNN) and Index Based Positioning for Keyword Search

Abstract

Conventional spatial queries, such as range seek and nearest neighbor retrieval, include only conditions on objects’ geometric properties. The proposed framework uses an productive calculation to find the accurate nearest neighbor based on the Euclidean separation for large-scale PC vision problems. We insert data focuses nonlinearly onto a low-dimensional space by straightforward calculations and demonstrate that the separation between two focuses in the implanted space is limited by the separation in the unique space. Instead of registering the separations in the high-dimensional unique space to find the nearest neighbor, a parcel of applicants are to be rejected based on the separations in the low-dimensional implanted space; due to this property, our calculation is appropriate for high-dimensional and large-scale problems. We too appear that our calculation is improved further by apportioning info vectors recursively. Opposite to most of existing quick nearest neighbor seek algorithms, our method reports the accurate nearest neighbor not an rough one and requires a exceptionally straightforward preparing with no modern data structures. We give the hypothetical examination of our calculation and assess its execution in manufactured and genuine data.