Tree-Based Graph Indexing for Fast kNN Queries

The k nearest neighbor (kNN) query on a graph is a problem to find k nodes having a shortest path distance from a user-specified query node in the graph. Graph indexing methods have the potential to achieve fast kNN queries and thus are promising approaches to handle large-scale graphs. However, those indexing approaches struggle to query kNN nodes on large-scale complex networks. This is because that complex networks generally have multiple shortest paths between specific two nodes, which incur redundant search costs in the indexing approaches. In this paper, we propose a novel graph indexing algorithm for fast kNN queries on complex networks. To overcome the aforementioned limitations, our algorithm generates a tree-based index from a graph so that it avoids to compute redundant paths during kNN queries. Our extensive experimental analysis on real-world graphs show that our algorithm achieves up to 146 times faster kNN queries than the state-of-the-art methods.