representing multiple dependencies in XES.

obordies25

Hi all,

I have a question about how represent multiple dependencies, as is represented in the Figure, in cvs format. My case is that the activity "c" depends as well on the realization of "b" as on the product generated by "a". I tried with several representations but results form heuristic or fuzzy mining where not consistent with these obtained by hand using pilots cases. Any help!!¿?

Thanks

JBuijs

Hi,

What do you mean by dependency?

A process model usually shows the direct dependencies, e.g. b depends on a and c depends on b. This implies hence c depends on a but this is not shown directly.

obordies25

Hi Joos, in my case the dependency between c and a have a different nature regarding the dependency between b and a, or c and b. The last two represent control flow dependencies (one event initiates onces the previous one ends), and the a-c dependency represent a dataflow. How i can represent in my logs this two nature. It's important to me to observe the interaction between this two kind od dependencies! 
thaks!!

JBuijs

Data dependencies are represented in the event log by adding the data values as attributes to the event. Dedicated event analysis techniques will then rediscover these.

I'll ask a colleague who worked on these techniques to explain in more detail

fmannhardt

Hi, you could use Data Petri nets and the techniques described here to model the data flow:
https://fmannhardt.de/blog/multi-perspective-process-explorer-mpe

The publication is here:
http://link.springer.com/article/10.1007/s00607-015-0441-1

I understand that you only want to check on the existence of a produced variable value (written in our terminology). That would be very easy. You just need to define your process as DPN, in which activity (transition in our terminology) writes a variable X. Since there is already a control flow dependency between A and C it is implicit that the value of X will have been written when C its executed. 

However you could also define a rule regarding the expected value of X. This is called a guard expression. For example, the guard X>1. This way, the model requires A to have written a value larger than 1 to X.