Indexing time-series databases for inverse queries

Abstract. In this paper we examine the problem of indexing time sequences in order to answer inverse queries. An inverse query computes all the time points at which the sequence contains values equal to the query value. The presented method is based on [7] in order to represent each time sequence with a few ranges of values, which are in fact one dimensional minimum bounding rectangles. We compare the proposed method with the IP-index which has been presented in [8] as an indexing mechanism for answering inverse queries. As it is shown, the proposed method outperforms the IP-index for very large time sequences.