Adaptive algorithms for balanced multidimensional clustering

Abstract

The G-K-D tree (generalized K-D tree) method aims at reducing the average number of data page accesses per query, but it ignores the cost of index search. The authors propose two adaptive algorithms that take into consideration both data page access cost and index page access cost. It attempts to find a minimum total cost. Experimental results indicate that the proposed algorithms are superior to the G-K-D tree method.<>