#SMEAR - SAX motif extraction and recognition#

written by Sebastian Steuer (steuers@cip.ifi.lmu.de)

##Requirements##

SMEAR depends on traits, traits.ui, and chaco.

traits is a type-system for python, traits.ui a MVC-framework and chaco is a chart plotting library. All three are made by enthought

• on debianesqes (ubuntu e.g.), try sudo apt-get install python-chaco. Go the hard way otherwise. (http://code.enthought.com/projects/chaco/docs/html/user_manual/installation.html)

##How to use##

• smear.py has help built-in. Try python src/smear.py -h

• plotter.py runs by itself and can:

  • load .arff/.wav/.sax data
  • select a slice of the sequence, select the parameters for the SAX-string generation
  • export the SAX-string as a .sax file for later comparison

• execute python src/smear.py -c SAXFILE to run a comparison of SAXFILE against all testdata and get a little evaluation (Precision, Recall, F1)

  • e.g python src/smear.py -c sax/e_1_word4.sax

you can also use the SAX-generator from an interactive python shell:

cd src
python
>>> import sax
>>> sax.saxify(range(100), epsilon = 20, word_size = 5, alphabet_size = 3)
'AABCC'

##Data##

The /td folder contains .arff-files for ten different exercises (1-10) performed by 5 different persons (a-e). Each file contains 10 iterations of the exercise.

##What does not work##

• non-unix-systems: hardcoded paths at a few places, Windows won't work.
• data with more or less than six dimensions, or in general all data that isn't exactly written in the format that SMEAR excepts.
• No unittests yet, tbd