SC²S Colloquium - May 16, 2018
|Date:||May 16, 2018|
|Time:||15:00 - 15:30|
Sebastian Roßner: Creating a Binary labeled Event Log for Process Mining using Semi-Supervised Log Pattern Detection and Exploration
Process Mining is a new technology using existing event log data to evaluate processes in their real run of events. An event log consists of multiple recorded events, with each event consisting at least of a case key, an activity name and a time stamp. Such events can be collected from any kind of process. With the help of process mining algorithms an as-is process model can be extracted from an event log and if there is an existing process model, a conformance check can be executed upon the recorded event log data. This leads to a list of violations of the target process. This list defines the potential optimization cases to create actions to get the as-is process as close to the target process as possible.
This thesis explains the basic assumptions on wanted and unwanted behavior and its definition in an event log. These assumptions are used to define approach for the definition of a labeled event log up to the point that we define a generic workflow model and algorithms to detect, classify and store wanted and unwanted behavior. We also provide a prototype implementation used for the analysis of two event logs from the Business Process Intelligence Challenge (BPIC) 2012 and 2017.
Keywords: Process Mining, Master Thesis Submission, Celonis