Detecting Process Concept Drifts from Event Logs

Traditional process discovery algorithms assume processes to be in a steady state. However, process models tend to be dynamic due to various factors, which has brought challenges such as change point detection, change localization and change process discovery. Existing techniques to identify change points are sensitive to parameters and the accuracy is not satisfactory. This paper proposes a novel approach to deal with such concept drift phenomenon. Event logs can be characterized by the relationships between activities, which motivates us to transform a log into a relation matrix. By detecting the always and never intervals in each row of the relation matrix, we obtain candidate change points for each relation. Finally, all the candidate change points are combined into an overall result. The approach is also able to localize the changes between different phases. Experiments on synthetic logs show that our approach is accurate and performs better than the state of the art in detecting sudden drift.

[1]  Wil M.P. van der Aalst,et al.  Genetic Process Mining , 2005, ICATPN.

[2]  Rafael Accorsi,et al.  Discovering Workflow Changes with Time-Based Trace Clustering , 2011, SIMPDA.

[3]  Mathias Weske,et al.  Efficient Consistency Measurement Based on Behavioral Profiles of Process Models , 2011, IEEE Transactions on Software Engineering.

[4]  Wil M. P. van der Aalst,et al.  The Application of Petri Nets to Workflow Management , 1998, J. Circuits Syst. Comput..

[5]  Jianmin Wang,et al.  A workflow net similarity measure based on transition adjacency relations , 2010, Comput. Ind..

[6]  Marlon Dumas,et al.  Fast and Accurate Business Process Drift Detection , 2015, BPM.

[7]  Olivier Cappé,et al.  Robust changepoint detection based on multivariate rank statistics , 2011, 2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).

[8]  Wil M. P. van der Aalst,et al.  Change Point Detection and Dealing with Gradual and Multi-order Dynamics in Process Mining , 2015, BIR.

[9]  M. Lavielle,et al.  Detection of multiple change-points in multivariate time series , 2006 .

[10]  Michèle Basseville,et al.  Detection of abrupt changes , 1993 .

[11]  Yoshinobu Kawahara,et al.  Change-Point Detection in Time-Series Data Based on Subspace Identification , 2007, Seventh IEEE International Conference on Data Mining (ICDM 2007).

[12]  F. Gustafsson.  The marginalized likelihood ratio test for detecting abrupt changes , 1996, IEEE Trans. Autom. Control..

[13]  V. Moskvina,et al.  An Algorithm Based on Singular Spectrum Analysis for Change-Point Detection , 2003 .

[14]  Josep Carmona,et al.  Online Techniques for Dealing with Concept Drift in Process Mining , 2012, IDA.

[15]  Evelina Lamma,et al.  Applying Inductive Logic Programming to Process Mining , 2007, ILP.

[16]  Arthur H. M. ter Hofstede,et al.  Characterizing Drift from Event Streams of Business Processes , 2017, CAiSE.

[17]  Shai Ben-David,et al.  Detecting Change in Data Streams , 2004, VLDB.

[18]  Akhil Kumar,et al.  ExRORU: A New Approach to Characterize the Behavioral Semantics of Process Models (Short Paper) , 2016, OTM Conferences.

[19]  Mykola Pechenizkiy,et al.  Dealing With Concept Drifts in Process Mining , 2014, IEEE Transactions on Neural Networks and Learning Systems.

[20]  Ricardo Seguel,et al.  Process Mining Manifesto , 2011, Business Process Management Workshops.

[21]  Matthias Weidlich,et al.  The 4C Spectrum of Fundamental Behavioral Relations for Concurrent Systems , 2014, Petri Nets.

[22]  Wil M. P. van der Aalst,et al.  Workflow mining: discovering process models from event logs , 2004, IEEE Transactions on Knowledge and Data Engineering.

[23]  Kenji Yamanishi,et al.  A unifying framework for detecting outliers and change points from time series , 2006 .

[24]  Olivier Capp'e,et al.  Homogeneity and change-point detection tests for multivariate data using rank statistics , 2011, 1107.1971.