Concept Drift Prediction Based on Event and Sensor Data Streams - GUI Widget
During runtime process execution can be subject to a variety of changes, e.g., exceptional situations or intended adaptations, resulting in so called process concept drifts.
In order to be able to react early, concepts drifts should be detected as early as possible or even be predicted. [SRM20] introduce a concept to predict process concept drifts and explain them based on drifts in associated sensor data streams. Sensor data streams may accompany process execution in different domains such as manufacturing (e.g., power) and medicine (e.g., blood pressure).
The core of the bachelor project is the concept presented in [SRM20] which uses the concept of process histories presented in [StRi18]. For the process mining algorithm(s) you can choose to implement it (them) yourself or to use existing implementations. For dynamic time warping, for example, the tslearn package can be used. The goal of the bachelor thesis is to design and implement a graphical user interface. Specifically, the presentation of the process execution information together with the sensor streams and the concept drifts is important.
Contact: bachelor.i17 [at] in.tum.de
Concept drift with dynamic time warping:
- Citation: [SRM20] Florian Stertz, Stefanie Rinderle-Ma, Juergen Mangler: Analyzing Process Concept Drifts Based on Sensor Event Streams During Runtime. BPM 2020: 202-219
- Springer: link.springer.com/chapter/10.1007%2F978-3-030-58666-9_12
- Author version: eprints.cs.univie.ac.at/6422/1/BPM2020_SMR.pdf
- Test data set: gruppe.wst.univie.ac.at/data/timesequence.zip
Process histories for concept drift detection:
- Citation: [StRi18] Florian Stertz, Stefanie Rinderle-Ma: Process Histories - Detecting and Representing Concept Drifts Based on Event Streams. OTM Conferences (1) 2018: 318-335
- Springer: link.springer.com/chapter/10.1007%2F978-3-030-02610-3_18
- Author version: eprints.cs.univie.ac.at/5830/1/StRi_CoopIS2018.pdf
Process mining algorithms implementation of your choice, e.g.:
- Tool: www.promtools.org
- Library: pm4py.fit.fraunhofer.de
Dynamic time warping:
- Citation: [Tav18] Tavenard, R.: tslearn documentation (2018)