BayCEER Workshop 2016
October 13, 2016
NW III, University of Bayreuth
SAX: A new tool for interactive haystack needle similarity search in time series
Christine Göhring1, Stefan Holzheu1, Oliver Archner1, Oleg Lobachev2
1 BayCEER IT-Group, University of Bayreuth
2 Applied Computer Science V, University of Bayreuth
2 Applied Computer Science V, University of Bayreuth
P 6.1 in Paving the way for research: Databases, instruments, networks & Open Session
BayCEER BayEOS database contains thousands of time series with over 200 million data points. Finding the most similar segments to some subsequence in the whole database, e.g. an interval of two weeks out of an air temperature time series, is no trivial task.
The presented web application addresses this challenge by using an adapted version of the SAX (Symbolic Aggregate Approximation) algorithm. By preprocessing the time series with SAX and using suitable searching parameters the web application provides results in less than one second.
Next to this poster there is a live demonstration of the web application where one can try out the interactive similarity search with a selection of air temperature time series.
Keywords: time series similarity search