Adaptive Query Processing in Point-Transformation Schemes

Point-transformation schemes transform multidimensional points to one-dimensional values so that conventional access methods can be used to index multidimensional points, enabling convenient integration of multidimensional query facilities into complex transactional database management systems. However, in high-dimensional data spaces, the computational complexity of the required query transformation is prohibitive. This paper proposes a nearoptimal solution based on our novel ideas of adaptive Z-ordering. The experimental results show that the proposed approach significantly improves the query performance of point-transformation schemes. The idea can easily be generalized to accommodate any hierarchical space-filling curve, not just the Zcurve.