Threshold Interval Indexing for Complicated Uncertain Data

Uncertain data is an increasingly prevalent topic in database research, given the advance of instruments which inherently generate uncertainty in their data. In particular, the problem of indexing uncertain data for range queries has received considerable attention. To efficiently process range queries, existing approaches mainly focus on reducing the number of disk I/Os. However, due to the inherent complexity of uncertain data, processing a range query may involve high computational cost in addition to the I/O cost. In this paper, we present a novel indexing strategy focusing on one-dimensional uncertain continuous data, called threshold interval indexing. Threshold interval indexing is able to balance between I/O cost and computational cost to achieve an optimal overall query performance. A key ingredient of the proposed indexing structure is a dynamic interval tree. The dynamic interval tree is much more resistant to skew than Rtrees, which are widely used in other indexing structures. We also present a more efficient version of our index, called the memoryloaded threshold interval index, which reduces the storage size so the primary tree can be loaded into memory. We perform experiments to demonstrate the effectiveness and efficiency of the proposed indexing strategy.