Analysis of Nearest Neighbor Query Performance in Multidimensional Index Structures

A frequently encountered type of query in geographic information systems and multimedia database systems is to find k nearest neighbors to a given point in a multidimensional space. Examples would be to find the nearest bus stop to a given location or to find some most similar images when an image is given. In this paper, we develop an analytic formula that estimates the performance for nearest neighbor queries and characterize the efficiency of multidimensional index structures for nearest neighbor queries. The developed formula can be used directly in the query optimizers and the characteristics of efficiency will become the basis for the design of the index structure. Experimental results show that our analytic formula is accurate within some acceptable error range. It is exhibited that the efficiency of the index structure depends on the storage utilization and the directory coverage of it.