TClass Overview

A summary of how TClass works is as follows:

• Look for sub-events - in the form specified by the user - in the training streams.

• Analyse the extracted sub-events and select sub-events that are important, typical or distinctive.

• For each of the training instances, look for these sub-events. The presence or absence of these temporal patterns can be used as attributes for propositional learning.

• Apply a propositional learner.

To illustrate how these work in practice, we return to the Tech Support domain.