Thoughts on:

      "Finding Patterns in Time Series: A Dynamic Programming Approach",
      Advances in Knowledge Discovery in Databases, 
      AAAI Press, 1996,
      Donald J. Berndt, 
      James Clifford,
      New York University

I do not have an electronic copy of this paper.
      
This paper presents a method for using dynamic time warping to find
patterns in temporal data. Patterns are represented as templates.
The time axis of a time series can be stretched or compressed to
match the template's. This makes the length of the time series negligible
in finding a match. By normalizing along the other axis, one can match
template patterns regardless of scale as well.
      
This seems relevant to our problem since we expect very different lengths
or number of values in our time series attributes. We have to have a 
way of comparing these time series of different lengths. While our
initial simplification of the problem using our "sustained" or 
"not sustained" characteristic precludes addressing the problem directly,
we may very well wish to enhance the system with more direct methods
of comparison.
      
After finding patterns it is the author's hope to mine association
rules from those patterns. This very closely parallels our own goal 
although our data set is more complex.