Replay a Log on Petri Net for Conformance Analysis

ewelterlich

Hey guys,

I have some questions regarding the Plug-in "Replay a Log on Petri Net for Conformance Analysis"

• Im pretty new to the field of process mining, when I use the plugin im not sure whether the tool uses Token Replay or Alignments for the check of conformance. As the most algorithms in this plugin use cost-based fitness, im guessing it uses Alignments. Is that correct?
• Does anybody know which formula is used to calculate the trace fitness or is there any paper where this is explained?

Thanks in advance for your help!

BR,

Edgar

hverbeek

Hi Edgar,

This plugin uses the alignments.

For the details on how fitness is computed using alignments, you could check the book "Process Mining: Data Science in Action "by Wil van der Aalst. In short, you first compute the best (cheapest) alignment for the *empty* trace. This provides you with the costs for the cheapest path through the model to reach the final marking using only model moves. If you then have a trace you want to compute fitness for, then you can easily compute the costs to replay this trace using only log moves. Together, the costs provide you with an upper bound for the costs for the trace, as you this combination of using model moves in the model and log moves in the trace will always work (providing you can reach the final marking in the model, of course). If you then find an optimal alignment for that trace, the trace fitness is computed as 1 - (the costs for the optimal alignment) / (the upper bound for the costs).

For log fitness, fitness is similar, but it is important tot note here that the log fitness is not defined as the average trace fitness. Instead, it is defined as 1 - (the sum of the costs for the optimal alignments of all traces) / (the sum of the upper bounds of all traces).

Kind regards,

Eric.

ewelterlich

Hey Eric,

thank you for the quick response!

KR,

Edgar