Efficient k Nearest Neighbor Query Processing on Public Transportation Network

Jiajia Li, Lingyun Zhang, Cancan Ni, Yunzhe An, Chuanyu Zong, Anzhen Zhang

Abstract: Existing electronic map apps, such as Baidu map, AutoNavi map, Google map, only provide the services “find k nearest POIs around me” and “give me the bus plan from s to d”. But they cannot answer “give me k POIs that I can reach earliest within one transfer by bus”. Such type of query is to find k nearest neighbors (kNN) on public transportation network (PTN), that is, returns k POIs whose arrival time are earliest starting from query location at some time and the number of required transfers is less than the user-specified constrain. Existing kNN algorithm on road network cannot be directly applied on PTN due to the more complex line combinations. Two algorithms TT-INE and TT-IER are proposed based on Incremental Network Expansion and Incremental Euclidean Restriction respectively. To address the loss of efficiency caused by the proliferation of alternative routes, a pruning strategy based on the route dominance relationship is designed. Moreover, to filter those stations who are not associated with POIs, a node compression method is also proposed, And a grid index is utilized to prune the POIs far away from the query based on the upper bound of arrival time. Extensive experiments have been conducted based on the real data of Beijing's PTN. Experimental results show that the TT-INE algorithm with pruning strategy is two orders of magnitude faster than the traditional incremental expansion method. When POIs are sparse, i.e. the percentage of POIs is 1 %, TT-IER is an order of magnitude faster than TT-INE.