A Research On The Dimension Reduction And Similarity Search For The Interval Time Series

Among all sort of data types in big data era, the time series has received increasingly attention. Technologies including dimension reduction, indexing and similarity searching on time series have been well studied. To meet some application’s requirement, we introduce a new data type-Interval Time Series which uses an interval rather than a single value to describe the measured value at each time point. In this paper we proposed the definition of the interval time series along with a dimension reduction algorithm and then implemented a similarity search system based on the R*tree.In this paper, we first proposed the definition of the interval time series, along with some analysis of its properties. Then an orthogonal transform based algorithm is proposed to reduce its dimensionality after a review of these classical dimension reduction techniques. We then implemented a similarity query system based on both the dimension reduction algorithm and the R*tree indexing structure. The experiments showed the algorithm and the system both works well.