Thoughts on:

"Identifying-and-Using-Patterns-in-Sequential-Data",

Philip Laird

NASA Ames Research Center,1993

I don't have an electronic copy of this paper.

This is a survey about learning from individual streams 
of data as they are generated. The main goal is predicting the next
symbol/value to appear in the sequence. While it might be useful to
identify patterns that can predict sequences and treat them as
characteristics of our time series attributes over which we
could mine association rules, it seems better to start with a
simpler characteristic in our initial investigations.

The approaches surveyed here might be applicable as a way to use
existing association rules. Imagine a system that monitors some
behavior. If, based on the stream so far a problem behavior is 
predicted, the rules that contain that problem might give some idea
of how to avoid it.