Notes on:

      "Adding Temporal Semantics to Association Rules",
      Chris P. Rainsford, John Roddick,
      1999, 
      3rd European Conference on Principles and Practice of Knowledge
      Discovery in Databases (PKDD'99)       
      
Diagrams and some discussion of this work can be seen in the
presentation Related located in the Document - Presentation section of
this web site.

This work "accommodates both point-based and interval-based models of
time simultaneously."

Good point: "Moreover, the algorithms only accommodate point-based
events and this restricts both the potential semantics of knowledge
that may be discovered and the data that can be learnt from." the
authors also point out that the work they cite are limited to finding
commonly occurring sequences rather than associations. The cited work
is: Agrawal, Srikant, "Mining Sequential Patterns", Heikki Mannila,
Hannu Toivonen, A. Inkeri Verkamo, "Discovery of frequent episodes in
event sequences", and Srikant, Agrawal, "Mining Sequential Patterns:
Generalizations and Performance Improvements".

Instances, in the example the authors use, clients, may be associated
with on-temporal properties. These are the same as regular attributes,
or items in ASAS.

Temporal attributes are defined as having points or intervals during
which the item or attribute was valid in the modeled domain.

Confidence is defined having two components. The first is the regular
confidence treating an association as if it had no temporal
component. The second is a percentage of the overall instances supporting the
association that also support a specified temporal relationship
between a single pair of temporal items. This is called the temporal
confidence of the pair given a specific relationship.

The process by which association s are found does not resemble my
own. The associations produced do not resemble my own. It may be able
to compare results but just superficially.