Whats the difference between a Quality Dimension, a Quality Metric, an Evaluation Metric, a Metric Parameter and the Quality of a Candidate?

A while ago I received the following question in my mailbox, and I figured I would reply through my blog, since it might be interesting to the ‘general public’ πŸ™‚

The question is as follows:

While I was reading chapter 4 of your PhD thesis, I wandered, what ‘s the difference between the following terms?

  1. Quality Dimension
  2. Quality Metric
  3. Evaluation Metric
  4. Metric Parameter
  5. Quality of Candidate

I think I’ve confused while embracing the different concepts of your ETM framework.

If you please, show me clear differences between them, as I think for somehow one can be an alternative to another in different context(s).

Interesting question, so let me briefly indicate the differences:

  1. Quality Dimension is an aspect a process model can be evaluated on. There are four common ones: replay fitness, precision, generalization and simplicity. Each quality dimension covers a different aspect of quality. Additional quality dimensions can be added, such as edit distance/similarity or configuration quality, as I show in my thesis.
  2. Quality Metric is a way to calculate a value expressing the quality of a particular process model, given an event log, on a particular quality dimension. An example are replay alignments as a metric to calculate the replay fitness quality dimension.
  3. Evaluation Metric is a more general notion of a quality metric, where a quality metric is specific for a quality dimension, an evaluation metric is any quality metric. I use this term in my thesis to present requirements on metrics in the context of the ETM evolutionary algorithm.
  4. Metric Parameter is a setting that can be required by a (quality/evaluation) metric. For example in alignments costs can be assigned for particular activities or for log or model move types, these settings are input to metric parameters.
  5. Quality of a candidate is the value assigned to a process model candidate on either a particular quality dimension (for instance replay fitness through alignments is 0.80), or the overall quality of a candidate process model.

I hope this clarifies some of the terminology used in my thesis, and mainly chapter 4 πŸ™‚

This article was first published on Blog of Joos Buijs: Whats the difference between a Quality Dimension, a Quality Metric, an Evaluation Metric, a Metric Parameter and the Quality of a Candidate?.

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to Top