I'm currently working on mining association rules from just the
kind of data you're asking about for my Master's thesis. What I've done is
specified those attributes which are temporal series as a string type in
arff. If you use a delimiter not used by arff between the values in the
series then Weka will read in the arff file just fine. Then in your
learning algorithm you can handle the temporal series as you wish. When
I've finish my thesis I'll be more than happy to share the source code and
other materials. Good luck.
Keith A. Pray kap(a)wpi.edu
WPI MS CS
WPI 1998 Class Board of Directors Chair
WPI ACM Executive Council
WPI Social Committee
ACM - UPE
On Tue, 19 Feb 2002, [iso-8859-1] Juan Andr�s Sastre De La Fuente wrote:
I would like to codify temporal series into arff format, but I have found a few
* How can I codify a multivariable serie?
* How about the series with a variable length?
* And what if the points don't appear in fixed intervals?
If someone can help me I would be very grateful for it.