Summary: | 碩士 === 國立中正大學 === 資訊工程研究所 === 101 === Skyline queries are an important technique for the search of multi-dimensi-
onal databases as useful knowledge is retrieved for decision making applica-
tions while observing user preferences. We address the new problem of the
potential skyline queries over dynamic objects with d dimensions. In con-
trast to most of the prior work, we focus on the issue of promoting a poten-
tial skyline, while handling both static and dynamic objects. In this paper,
we propose an algorithm termed DyE-SPOT: a Dynamic Efficient Search for
Potential Skylines, and design an efficient promotion technique for potential
skylines over dynamic objects, by utilizing a pre-computed second-ordered
skyline set as a candidate set for the promotion. The second-ordered skyline
facilitates an efficient and incremental promotion technique to retrieve a set
of top k potential skyline ranked by promotion cost. With the knowledge of
the second-order skyline set as a candidate set for promotion, our approach
efficiently finds the potential skyline points ranked by their promotion cost in
ascending order. Finally, we design a cost model to evaluate the promotion
cost which is eventually utilized in a pruning method to avoid scanning the
entire data set for promotion. We compare our method to the existing tech-
niques and the results demonstrate better performance. Extensive experiment
results show that our method on dynamic datasets are both efficient and scal-
able.
Keywords: Multi-dimensional Databases, Spatial-temporal Databases, Dy-
namic skyline queries, Potential skyline search
|