Eamonn
Keogh,
Jessica Lin, Ada Fu
This webpage provides some extra details about time series discords. The
first paper on time series discords was:
E. Keogh, J. Lin and A. Fu (2005). HOT SAX: Efficiently Finding the Most Unusual Time Series Subsequence. In The Fifth IEEE International Conference on Data Mining.
- Click here for a slightly longer version
of the paper. This 10 page version has more experiments, more references and
more detailed explanations.
- Click here to download a PowerPoint presentation many with extra discord
experiments (PDF version)
- Click here or here
to learn
more about SAX, and to get free SAX code.
Datasets
Important note: the figure numbers listed below point to the
figures in the longer version of the paper.
- Click here to download all the ECG datasets
used in slides 2 to 8.
- Click here to download the video surveillance
dataset used in slide 11.
- Click here to download the Space Shuttle dataset
used in slide 13. (Fig 6 in the paper)
- Click here to download the Space Shuttle dataset
used in slide 14.(Fig 7 in the paper)
- Click here to download the Space Shuttle dataset
used in slide 15.
- Click here to download the ECG
dataset used in slide 17. (Fig 12 in the paper)
- Click here to download the ECG dataset
used in slide 18. (Fig 13/14 in the paper)
- Click here to download the ECG dataset used
in slide 19. (Fig 11 in the paper)
- Click here to download the respiration dataset
used in slide 20. (Fig 9 in the paper)
- Click here to download the respiration dataset
used in slide 21. (Fig 10 in the paper)
- Click here to download the power demand
dataset used in slide 22. Fig 15 in the paper)
- Click here to download the ECG dataset used
in slide 30. (Fig 1 in the paper)
You can obtain all the above datasets, plus hundreds more by requesting a
free CD-rom from the UCR Time Series Data Mining Archive.