For my master thesis i am investigating the novel work conducted in the Petra project and would very much appreciate some input from any and all who were either connected to the project, or have experiences with the Petra plugin. Petra being a acronym for "Process model based Extensible Toolset for Redesign and Analysis"
My primary intention is to use Petra to conduct process enhancement by exploring different model alternatives. It will be used to "what if" questions about varying resource availability, model composition, and inbound traffic.
From my understanding, the Petra project answered many of these "what if" questions by use of model simulation and analysis. The Petra plugin(s) being the primary tool that allowed for such analysis.
While the project has ended, the project details are still available on https://www.win.tue.nl/coselog/wiki/start
Without doing the work any injustice i will attempt to describe its intention. Petra is a toolset that allows for the execution and analysis of a "configurable model"
. This model is represented as a process tree and is the product of the union of a solution space that consist of a collection of similar, but deviating models. This solution space is described as "process families", being that they are closely related.
The user can then change properties on the configurable process model via the GUI in the plugin(i assume). This allows for the creation of a series of process models that correspond to the respective "settings" of configurable process models. These are then offered through one of two interfaces. The outbound interface offers up one of the models in the series, the model will then have to be transformed into a formalism and executed in a analytical tool, the results of the analysis are then send to the inbound interface where they are used to add annotations to the same process model. In the end you have a collection of annotated process models and you can locate the "best" model according to your perspective.
With my lack of experience, i am tempted to describe this approach as a "brute force" attempt at finding the best suited process model depending on your need. This is because the plugin produces a complete set of process models, i.e. the set should be comprised of all possible alterations of the input model.
While computationally expensive, it seems and is a good way of finding exactly the model that suits your need.
My intention being to propose and execute a opinionated implementation of the toolset. While the toolset supports the use of other formalism and analytical tools, i will use CPNs and CPN Tools/Access 2.0, just as they did in the project.
However, i have come to a bit of a halt, as i often do. I am finding it difficult to locate sources on exactly how process trees are created from event logs and what format they are represented in. What data to these trees encode?
What input does the petra plugin take and how is this represented.
I am finding it hard to understand how we can go from a event log (.XES) to a process tree/configurable process model that can be used as input in the plugin. This is seemingly not mentioned in any of the papers..
Explicit examples are very welcome!
As always, any and all input is much appreciated!