Dimensionality reduction techniques for streaming time series: a new symbolic approach

A growing number of applications generates massive streams of data which are on-line collected and potentially unbounded in size. To cope with the high dimensionality of data, several strategies for dimensionality reduction have been proposed. In this paper we introduce a new approach to represent an append only data stream into a reduced space. The main aim is to transform a real valued data stream into a string of symbols. The string includes a level component and a shape component allowing to get a better representation of data while maintaining a strong compression ratio.