A Practical Application of Predictive Process Monitoring in Cloud Computation at <<Company Disclosed>>

Kontakt

Name

Sebastiaan J. van Zelst

Wissenschaftlicher Mitarbeiter - Fraunhofer FIT

Telefon

work
+49 241 80 21926

E-Mail

E-Mail
 

Student: Xuan Zhang

Title: A Practical Application of Predictive Process Monitoring in Cloud Computation at <<Company Disclosed>>

Supervisor: Dr. Sebastiaan J. van Zelst

1st Examiner: Prof. Wil van der Aalst

2nd Examiner: Prof. Wolfgang Prinz

Summary

Today’s business processes are supported by a variety of information systems, and every executed activity leaves a digital footprint in the form of event data. Process mining leverages event data to discover, monitor, and improve operational processes, and helps enterprises to achieve business intelligence. Traditional process mining techniques utilize the existing event data to analyze and monitor the business process, and are often categorized as either: process discovery, conformance checking, or process enhancement techniques. Meanwhile, a growing number of studies and industrial interest is shifting to providing continuous monitoring and automated analysis for the ongoing process. Identifying inefficiencies in real time and integrating the analysis seamlessly in operation can bring significant business values to the organization, for it allows operational experts to take immediate preventive actions proactively, to avoid undesired outcomes. A general framework for such action-oriented applications has been proposed, however, a practical instantiation faces more challenges in operation. In this thesis, we implement a real-life case study of the action-oriented process monitoring in collaboration with the Global Service department at <<Company Disclosed>>. We study the feasibility and applicability of the architectural framework, and extend it with predictive process mining, where we emphasize on the effect of concept drift detection to cope with the increased impact of behavioural changes in the process. Moreover, we present a quantitative evaluation on the prediction performance and qualitatively evaluate the participants feedback on the case study with <<Tool Vendor Disclosed>>. Additionally, we propose a set of best-practice guidelines for future implementation.

Note

This work was performed in collaboration with an industrial partner. Therefore, the final thesis is not shared publicly. A brief project description is made available. For more information, contact Dr. Sebastiaan J. van Zelst