Efficient Filtering Query Indexing in Data Stream

Filtering queries are widely used in data stream applications. As more and more filtering queries are registered in high-speed data stream management system, the processing efficiency becomes crucial. This paper presents an efficient query index structure based on decision tree. The index structure makes full use of predicate indices on single attributes, as well as the conjunction relationship between predicates in a single query. It is easy to integrate various predicate indices into this structure. How to select dividing attributes during construction is crucial to the performance of the index tree. Two dividing attribute selection algorithms are described. One is based on information gain (IG) and the other is based on estimated time cost (ETC). The latter takes some sample tuples as a training data set and is able to build more efficient trees. Our experiments demonstrate that.

• Type: Conference
• Year: 2006
• Where: WISE
• Authors: Ying Wang, Shuo Bai, Jianlong Tan, Li Guo
• Keywords: Data Stream, Filtering Queries, Internet Technology, Predicate Indices, WISE 2006