ProM 6.10 conformance checking

rbrandes

Hello there,

I am still quite new to the world and currently trying to do some conformance checking with ProM 6.10.

I have process model (only a .pdf file though) and a large set of event logs which differ quite a lot. I looked for a petri net model generator to derive a petri net model from my .pdf process model but could not find a fitting ressource. 
I also tried generating a petri net model using some ProM plugins for process discovery which did not fit enough to be used.

So my question is: does anyone has experience with a plugin for conformance checking with ProM 6.10 and how can I generate a fitting process model with a) via software/prom or b) via a petri net generator.
The process itself is not that complicated, its more of a sequence of activities with 2-3 OR splits.

Thank you and have a nice day

Richard

hverbeek

Hello,

You can use CPN Tools (see http://cpntools.org/) for modeling your Petri net, and then import it using the "Import Petri net from CPNXML file" plugin in ProM 6.10. If so, please provide every place in your model with a proper non-empty name. CPN Tools can handle empty place names, but the importer plugin in ProM 6.10 can not.

By default, CPN Tools will use the UNIT type for every place and the "1`()" expression for every arc, which is fine. If you want to make a transition silent (that is, if it is not related to any event in your logs), please provide it with a black background color.

ProM 6.10 also has a "CPNXML export (Petri net)" export plugin that allows you to save a Petri net to a .cpn file that can be imported into CPN Tools. Using both plugins, you can go back and forth between ProM 6.10 and CPN Tools.

If you have an event log, then you can create a fitting Petri net using, for example, the "Mine Petri net with Inductive Miner" plugin (selecting the "Inductive Miner (IM)" variant). This will discover a Petri net where every trace fits, but it may not be too precise. To get more precise (but possibly unfitting) results, keep the "Inductive Miner - infrequent (IMf)" variant, and play with the noise threshold.

Kind regards,

Eric.