Convert process tree to Petri net

JBuijs

This discussion was created from comments split from: How to compare process models?.

janan@yahoo.com

Hi dear Buijs

can I convert ETM to Petri net form in ProM 6 , 6.1, 6.2,...?  I tried to convert ETM to Petri net form in Prom6 but I couldn't find it.

JBuijs

Hi Janan, This should be possible in ProM 6.5 and 6.6.

janan@yahoo.com

but I couldn't find it. is it possible send me its link? Thank you

JBuijs

Hi Janan,

Please see screenshot in ProM 6.5.1, but this should also be there in ProM 6.6:

Screenshot: http://www.win.tue.nl/~jbuijs/files/PTtoPNscreenshot.png

janan@yahoo.com

Dear JBuijs

Thank you for you replaying.

janan@yahoo.com

Dear JBuijs

regarding your helping I could convert process tree obtained ETM Miner to Petri-net. but Petri-net converted is not able to show fitness, precision and simplicity. is there way to show 4 metrics of conformance checking show for petri net in ProM6.5.1?

janan@yahoo.com

Dear JBuijs

in order to compare  my proposed method with other approaches, 4 evaluation metrics fitness, precision, simplicity and generalization are used. then 3 compression approaches are run by Prom5.2 and ETM is run by Prom6.5.1. there is a question that whether the formula used in both ProM version is the same. For example, to measure fitness in Prom 5.2  has been used log replay or precision is used behavioral appropriates and  Simplicity is used base on Rozinat and van der Aalst (2008).   

JBuijs

Dear Janan,

The short answer: the metrics used in ProM 5.2, ProM 6 for Petri nets, and in ProM 6 for process trees all three are different for all 4 quality dimensions.

Longer answer:
In ProM 6 we use alignments for replay fitness, and precision and generalization. This technique did not exist in ProM 5.
For process trees we implemented a specific alignment algorithm that is more efficient because it uses process tree characteristics for this. Hence, the idea is the same, but the specific algorithm is different.

Solution: use the Petri net replay using alignments for conformance, precision and generalization in ProM 6. Export the Petri nets discovered in ProM 5.2 and import them again in ProM 6. Also translate the process tree to a Petri net and apply the same analysis techniques.

Apply the alignment ('replay using conformance' or something similar) plug-in for this, but please add an initial and final marking to the Petri net for this.

janan@yahoo.com

thank you. if I want to use metrics used in Prom5.2, can I export the Petri net discovered of ETM miner in ProM 6.2.1 and then import it in Prom 5.2?

JBuijs

I think you could (the best answer is to 'just try' ). But I would strongly recommend using ProM 6 for this as the techniques are more recent and more robust.

For instance, if you write a scientific paper and I would need to review it, for me this would be a strong argument to vote to reject the paper for publication.

Hope this helps.

janan@yahoo.com

Dear JBuijs

thank you. when I covert process model discovered of ETM to Petri net, how can I measure 4 metrics of conformance checking?

JBuijs

Hi Janan,

Although I'm willing to help, I'm also assuming that you searched before you asked.

In short: replay fitness, precision and generalization can all be calculated using alignments. Simplicity can not directly be measured on the Petri net, but there are several metrics that are easy to manually calculate. This should give you enough keywords to search this forum and process mining literature.

janan@yahoo.com

Dear Buijs

Thank you so much. I could get expected result with your helping.

janan@yahoo.com

Dear Buijs

I could convert ETM to Petri net but Petrinet obtained is with so many invisible task. after converting, I chose a log and that petri net  and  apply "replay a log on Petri net for conformance checking" as you mentioned. but according to adriansyah thesis, show finesse on move log and fitness on model.  is there any plug in that shows value of precision, fitness, generalization like process tree obtained of ETM?

Thank you 

JBuijs

Hi Janan,

The translation could indeed contain many silent transitions. You could see how many are removed by executing one of the 'reduce Petri net' (not sure of the exact name).

I'm not sure what you mean with 'but according to adriansyah thesis'.

You can use the alignment to calculate replay fitness. There are also plug-ins that calculate precision and generalization, did you search for these in ProM?

janan@yahoo.com

 Hi Buijs

yes I found this plug-in  prom6.6. I had problem with many invisible task after converting. but I could remove them. I am so thanks full to replay me.

janan@yahoo.com

Dear Buijs

firstly, I am grateful to  to you for the patient guidance. you help me more to finish my evaluation results in my thesis.  secondly, I have a petri-net model by hand and tried to draw it by ( visio, graphvis, concept draw project) with vsd, vdx format. but prom couldn't open it. is there any way to draw petri-net and then import in prom?

JBuijs

Hi Janan,

I would recommend the Petri net tool WoPeD:
http://woped.dhbw-karlsruhe.de/woped/