Parallel computing of KNN Query in road network based on MapReduce

Abstract

K-nearest neighbor (KNN) query algorithm based on road network plays an important role in location based service, which had been widely used in intelligent transportation, roadside assistance and other fields. However, as road network density increases and the number of points of interest increases, query efficiency decreases sharply.In order to improve the query efficiency, we adopted the MapReduce parallel computing framework to complete the query of K neighbor moving objects by designing Map, Reduce, Combiner and other functions. Before the start of the query, the road network was divided into pieces, and each fragment was calculated. The final K-nearest neighbor moving objects were obtained by aggregating the calculated results of each slice to realize the parallel optimization of KNN algorithm based on road network. The experimental results showed that the performance of parallel KNN algorithm based on MapReduce was better than that of serial KNN query algorithm in large-scale road network environment and the larger K value of query request.