Spatial network k-Nearest neighbor: A survey and future directives

Abstract

Nearest neighbor algorithms play many roles in our daily lives. From facial recognition to networking applications, many of these are constantly improved for faster processing time and reliable memory management. There are many types of nearest neighbor algorithms. One of them is called k-nearest neighbor (k- NN), a technique that helps to find number of k closest objects from a user location within a specified range of area. k- NN road network algorithm studies have been through various query performance discussions. Each algorithm is usually judged based on query time over few selected parameters which are; number of k, network density and network size. Many studies have claimed different opinions over their techniques and with many results to prove better query performance than others. However, among these techniques, which k- NN road network algorithm has the highest rate of query performance based on the selected parameters? In this paper, reviews on several k nearest neighbor algorithms were made through series of journal extractions and experimentation in order to identify the algorithm that achieves highest query performance. It was found that with the experimentation method, we can identify not only the algorithm’s performance, but also its design flaws and possible future improvement. All methods were tested with some parameters such as varying number of k, road network density and network size. With the results collected, Incremental Expansion Restriction – Pruned Highway Labeling method (IER- PHL) proves to have the best query performance than other methods for most cases.