Dataset Generation for Lifecycle Process Mining
Project Description
Many organizations run processes that never truly finish: a software bug gets reworked and reopened, a patient with a chronic disease relapse and returns to treatment, a compliance check must be renewed repeatedly. These lifecycle processes look nothing like the one-directional, goal-driven processes that most process analysis tools are built to study and that public benchmark datasets such as the BPI Challenge logs describe. Because existing benchmark logs do not capture this behavior, methods for analyzing lifecycle processes cannot be systematically compared or validated, and theories of lifecycle behavior remain speculative for lack of real data to test them against. Building a purpose-built dataset is therefore an essential step for this emerging field. This project develops a pipeline that extracts lifecycle-aware event logs from real-world software repositories, which evolving histories of code, issues, and releases make lifecycle behavior directly observable. Both the extraction framework and the resulting datasets, comprising millions of events, will be published openly, so the scientific community can reuse the pipeline on new sources and build directly on the released data.
Relevance to Liechtenstein
The project develops an openly available dataset and extraction framework for lifecycle process mining at the University of Liechtenstein. Publicly released datasets are widely reused in process mining research, so this contributes to the international visibility of Liechtenstein-based research in this community. The pipeline, dataset, and characterization tooling will be released publicly; researchers, students, and interested organizations can reuse and extend them, so the results remain useful beyond the four-month funding period.
Financial services and industrial firms in Liechtenstein operate lifecycle-driven processes, such as client and product lifecycles or long-term software and product evolution. The methodology developed here provides a first step toward analyzing such evolving processes with process mining methods. Similar potential exists in healthcare, where long-term patient trajectories share the same characteristics.
Financial services and industrial firms in Liechtenstein operate lifecycle-driven processes, such as client and product lifecycles or long-term software and product evolution. The methodology developed here provides a first step toward analyzing such evolving processes with process mining methods. Similar potential exists in healthcare, where long-term patient trajectories share the same characteristics.
