How can I create a petri net with 100% precision?

ChrisC

I have created a simple event log to test the ‘Mine Petri Net with Inductive Miner’ plug in.  The event log has the following four cases:
A, B, E, F

A, C, E, F

A, D, E, F

A, F

I would like to create a petri net with 100% precision, so I would expect the diagram to look something like this:

Instead the plug-in gives me the following diagram:

The silent transitions are creating additional states.  It appears that the plug-in is under-fitting the data.  The above petri net allows for cases which are not found in the event log (such as A, B, F and A, E, F).  I have tried adjusting the noise threshold in the plug-in settings but this does not solve the problem.  How can I create the desired petri net with 100% precision?  Thanks.

kfgz087

Hi ChrisC,

The model discovered by Inductive Miner cannot guarantee a100% precision, and the noise threshold cannot help because all traces are observed only once.

I think you can add a silent transition on the bottom edge in your designed Petri net, which can get a 100% precision.

sjjleemans

As ABC & E are always skipped together, this is a case that IM should be able to handle. I'll have a look...

sjjleemans

Yup, that was a bug. It should have been fixed now, and appear in the Nightly build tomorrow.

Thanks for reporting!

ChrisC

Thank you kfgz087 and sjjleemans for your responses.

After downloading the Nightly Build and using Version 6.8.462 of the InductiveMiner (was previously using 6.7.391), the ‘Mine Petri Net with Inductive Miner’ plug-in can now create the following petri net:



This diagram has perfect precision. Thank you sjjleemans for updating the plug-in.