Data-aware Synthetic Log Generation for Declarative Process Models
In Business Process Management, process mining is a class of techniques for learning process structure from an execution log. This structure is represented as a process model: either procedural or declarative. Examples of declarative languages are Declare, DPIL and DCR Graphs. In order to test and improve process mining algorithms a lot of logs with different parameters are required, and it is not always possible to get enough real logs. And this is where artificial logs are useful. There exist techniques for log generation from DPIL and declare-based models. But there are no tools for generating logs from MP-Declare – multiperspective version of Declare with data support. This thesis introduces an approach to log generation from MP-Declare models using two different model checkers: Alloy and NuSMV. In order to improve performance, we applied optimization to baseline approaches available in the literature. All of the discussed techniques are implemented and tested using existing conformance checking tools and our tests. To evaluate performance of our generators and compare them with existing ones, we measured time required for generating log and how it changes with different parameters and models. We also designed several metrics for computing log variability, and applied them to reviewed generators.
Graduation Thesis language
Graduation Thesis type
Master - Software Engineering
Fabrizio Maria Maggi, Chiara Di Francescomarino, Chiara Ghidini