MaxBRkNN Queries for Streaming Geo-Data

Abstract

The problem of maximizing bichromatic reverse k nearest neighbor queries (MaxBR$k$NN) has been extensively studied in spatial databases, where given a set of facilities and a set of customers, a MaxBR$k$NN query returns a region to establish a new facility p such that p is a $k$NN of the maximum number of customers. In the literature, current solutions for MaxBR$k$NN queries are predominantly static. However, there are numerous applications for dynamic variations of these queries, including advertisements and resource reallocation based on streaming customer locations via social media check-ins, or GPS location updates from mobile devices. In this paper, we address the problem of continuous MaxBR$k$NN queries for streaming objects (customers). As customer data can arrive at a very high rate, we adopt two different models for recency information (sliding windows and micro-batching). We propose an efficient solution where results are incrementally updated by reusing computations from the previous result. We present a safe interval to reduce the number of computations for the new objects, and prune the objects that cannot affect the result. We perform extensive experiments on datasets integrated from four different real-life data sources, and demonstrate the efficiency of our solution by rigorously comparing how different properties of the datasets can affect the performance.