ML Wiki
Machine Learning Wiki - A collection of ML concepts, algorithms, and resources.

Process Conformance

Workflow Conformance Checking

This procedure gives the answer on the following questions

  • does our model conform to what actually happens?
  • does what actually happens conform to the model?
  • should we change our workflow to better fit the reality?

So it’s

  • what actually happens vs
  • what we think happens

Image

Play In

Suppose we have some logs that were captured during the process execution

  • we can ‘'’play in’’’ these sequences of actions on the model
  • and see if any of them fails
  • if yes - then we have no conformance

Image

This is very helpful in

  • finding tasks that are not automated and performed outside of the engine
  • (sometimes we turn off the workflow engine and add some exceptional cases manually)
  • are there lots of these exceptions? then we should add them to the model

Simulation (Play Out)

We can try to simulate the execution of our model

  • to see what can change if we add or remove something
  • are there any bottlenecks that have appeared after the change?

To do this we need to define the environment of execution

  • need to have the process itself
  • for each activity
    • times and priorities
    • statistics on how often a certain branch is followed
  • frequency of incoming cases
  • the number of resources that can handle the cases

So given the statistics

  • a some kind of random walk process is executed
  • it basically traverses the Reachability Graph of a model
  • and emulates the delays, etc

Image

After the execution we can see:

  • avg execution time
  • resource utilization
  • etc

Sources