SBT-forest, an indexing approach for specialized binary tree

In our previous work, a time series representation framework, Specialized Binary Tree (SB-Tree) has been proposed for representing the stock time series data effectively and efficiently. By putting a set of SB-Trees together, a time series database is formed while we termed it as a Specialized Binary Tree-Forest (i.e. SBT-Forest). By manipulating the SBT-Forest, different time series query and mining processes can be facilitated. However, the major challenge is how to locate a SB-Tree in the forest efficiently. Therefore, the development of an indexing approach for the SB-Trees is of fundamental importance for maintaining an acceptable speed for query. In this paper, a time series indexing approach, based on transforming the SB-Trees to symbol strings first and then indexing the symbol strings by a trie data structure, is proposed. The proposed approach is efficient and effective as well. As demonstrated in the experiments, the proposed approach speeds up the time series query process. The proposed approach can handle the problem of updating new entries to the database without any difficulty.