Efficient k-Nearest neighbor search for static queries over high speed time-series streams

Abstract

In this paper, we propose a solution to the multi-step k-nearest neighbor (k-NN) search. The method is the reduced tolerance-based k-NN search for static queries in streaming time-series. A multi-scale filtering technique combined with a multi-resolution index structure is used in the method. We compare the proposed method to the traditional multi-step k-NN search in terms of the CPU search time and the number of distance function calls in the postprocessing step. The results reveal that the reduced tolerance-based k-NN search outperforms the traditional k-NN search. Besides, applying multithreading to the proposed method enables the system to have a fast response to high speed time-series streams for the k-NN search of static queries.