Improving Sort-Tile-Recusive algorithm for R-tree packing in indexing time series

Abstract

The Sort-Tile-Recursive (STR) algorithm is a simple and efficient bulk-loading method for spatial or multidimensional data management using R-tree. In this paper, we put forward an approach to improve the STR algorithm for packing R-trees in indexing time series by some strategies choosing coordinates to partition spatial objects into nodes of R-trees. Every strategy has its own method to connect ends of consecutive runs into a suboptimum space-filling curve. We will compare the proposed approach with previous works in terms of space storing the index structure and runtime for range search on R-trees. Extensive experiments are carried out on many streaming time series datasets to evaluate our improved STR methods and previous methods unbiasedly and precisely. The experimental results show that the improved STR methods outperform previous methods.