The Area Code Tree for Approximate Nearest Neighbour Search in Dense Point Sets

In this paper, we present an evaluation of nearest neighbour searching using the Area Code tree. The Area Code tree is a trie-type structure that organizes area code representations of each point of interest (POI) in a data set. This data structure provides a fast method for locating an actual or approximate nearest neighbour POI for a query point. We first summarize the area code generation, insertion (used in overall construction) and searching approaches. Then, we evaluate the contraction, insertion, searching, and accuracy of the Area Code tree, with some of the evaluation involving the comparison versus a basic benchmark brute force approach. We find that when the Area Code tree is used for locating an approximate nearest neighbour, that low constant-time search is achieved, and in denser POI sets, higher accuracy is achieved. This ultimately makes the Area Code tree a strong candidate for approximate continuous nearest neighbour processing for location-based services. keywords: nearest neighbour queries, spatial access methods, location-based services