TY - GEN
T1 - Time drift detection in process mining
AU - Che, Haiying
AU - Machu, Quentin
AU - Zhou, Yangguang
N1 - Publisher Copyright:
© 2016 Taylor & Francis Group, London.
PY - 2016
Y1 - 2016
N2 - Currently, most of the information systems can record the tracking information and logs, this helps people to know the performance of the process execution. Process Mining techniques allow knowledge extractions such as model discovery, conformance checks and process improvements to take place. Processes are subject to various changes during their execution, for instance, a change in structure may occur when a new regulation comes into force and imposes some change, or may happen under the influence of seasonal effects, natural disasters etc. For many industries, time is a crucial factor in most cases equal to efficiency and profitability. Thus, this research paper presents an approach for detecting time-related changes. Our method extracts time-related characteristics from processes and then compares all of them together by using statistical hypothesis tests in different successive populations. Such a method could not only allow accurate detection when some parts of the processes started to have abnormal behavior: longer or shorter but also enable identification of which parts are involved. Based on the proposed approach in this paper, a ProM6 plug-in is implemented and tested. Further, synthetic data is used to do the experiment, finally, the results are explained and discussed.
AB - Currently, most of the information systems can record the tracking information and logs, this helps people to know the performance of the process execution. Process Mining techniques allow knowledge extractions such as model discovery, conformance checks and process improvements to take place. Processes are subject to various changes during their execution, for instance, a change in structure may occur when a new regulation comes into force and imposes some change, or may happen under the influence of seasonal effects, natural disasters etc. For many industries, time is a crucial factor in most cases equal to efficiency and profitability. Thus, this research paper presents an approach for detecting time-related changes. Our method extracts time-related characteristics from processes and then compares all of them together by using statistical hypothesis tests in different successive populations. Such a method could not only allow accurate detection when some parts of the processes started to have abnormal behavior: longer or shorter but also enable identification of which parts are involved. Based on the proposed approach in this paper, a ProM6 plug-in is implemented and tested. Further, synthetic data is used to do the experiment, finally, the results are explained and discussed.
UR - http://www.scopus.com/inward/record.url?scp=85016802082&partnerID=8YFLogxK
U2 - 10.1201/b21308-15
DO - 10.1201/b21308-15
M3 - Conference contribution
AN - SCOPUS:85016802082
SN - 9781138028814
T3 - Signal and Information Processing, Networking and Computers - Proceedings of the 1st International Congress on Signal and Information Processing, Networking and Computers, ICSINC 2015
SP - 99
EP - 108
BT - Signal and Information Processing, Networking and Computers - Proceedings of the 1st International Congress on Signal and Information Processing, Networking and Computers, ICSINC 2015
A2 - Chen, Na
A2 - Huang, Tingting
PB - CRC Press/Balkema
T2 - 1st International Congress on Signal and Information Processing, Networking and Computers, ICSINC 2015
Y2 - 17 October 2016 through 18 October 2016
ER -