Similarity Search in Recent Biased Data Stream with Sliding Windows

Abstract

Similarity search in time series databases is an important research direction. Research in this field has focused on the development of effective transformation techniques, the application of dimensionality reduction methods and the design of efficient indexing schemes. In the case where time series are continuously updated with new values (streaming time series), the similarity problem becomes even more difficult to solve, since we must take into consideration the new values of the series. The challenge in a streaming database is to provide efficient algorithms and access methods for query processing, taking into consideration the fact that the database changes continuously as new data become available. Traditional access methods that continuously update the data are considered inappropriate, due to significant update costs. To attack the problem, significant research has been performed towards the development of effective and efficient methods for streaming time series processing. In this paper, we introduce the most important issues concerning similarity search in recent biased time series databases, presenting an efficient technique for similarity query processing in streaming time series using sliding windows. The proposed method called as adaptive stream processing is based on an incremental computation of Discrete wavelet transform which is used as a feature extraction method. In order to prove the efficiency of the proposed method, experiments have been performed for range query and k-nearest neighbor query on real-life data sets. The results have shown that the adaptive stream processing method exhibit consistently better performance in comparison to previously proposed approaches.