Note on:

     "Rule Discovery from Time Series"
     Gautam Das, King-Ip Lin, Heikki Mannila, 
     Gopal Renganathan, Padhraic Smyth,
     In  Proc. of the 4th Int. Conf. on 
     Knowledge Discovery and Data Mining, 
     pages 16-22. AAAI Press, 1998.

The input data set for the algorithm presented in this paper is a
numeric sequence or set of numeric sequences. All of these sequences
share the same time line. 

Each sequence is partitioned into windows, the size of which is a user
specified parameter. The windows are clustered based on similarity and
assigned an identifier. The original sequences are then partitioned
and these partitions are labeled with the cluster identifiers.

Apriori Sets And Sequences relies on events being discovered in the
sequences as a preprocessing step. This is more flexible allowing any
event detection method to be used.

A whole data set here could be translated into a single instance for
Apriori Sets And Sequences.

Rules found from this data are in the form A => B where B follows A in
T time units. The confidence of a rule is the fraction of occurrences
of A that are followed by B within T units.

Apriori Sets And Sequences finds more general rules that include any
temporal relationship between A and B. The measure of confidence used
in Apriori Sets And Sequences is similar in notion.