Deep Learning Process Prediction with Discrete and Continuous Data Features

Process prediction is a well known method to support participants in performing business processes. These methods use event logs of executed cases as a knowledge base to make predictions for running instances. A range of such techniques have been proposed for different tasks, e.g., for predicting the next activity or the remaining time of a running instance. Neural networks with Long Short-Term Memory architectures have turned out to be highly customizable and precise in predicting the next activity in a running case. Current research, however, focuses on the prediction of future activities using activity labels and resource information while further event log information, in particular discrete and continuous event data is neglected. In this paper, we show how prediction accuracy can significantly be improved by incorporating event data attributes. We regard this extension of conventional algorithms as a substantial contribution to the field of activity prediction. The new approach has been validated with a recent real-life event log.

[1]  Jana-Rebecca Rehse,et al.  Predicting process behaviour using deep learning , 2016, Decis. Support Syst..

[2]  Boudewijn F. van Dongen,et al.  Cycle Time Prediction: When Will This Case Finally Be Finished? , 2008, OTM Conferences.

[3]  Jürgen Schmidhuber,et al.  Deep learning in neural networks: An overview , 2014, Neural Networks.

[4]  Moe Thandar Wynn,et al.  Predicting Deadline Transgressions Using Event Logs , 2012, Business Process Management Workshops.

[5]  Wil M. P. van der Aalst,et al.  Time prediction based on process mining , 2011, Inf. Syst..

[6]  Boudewijn F. van Dongen,et al.  XES, XESame, and ProM 6 , 2010, CAiSE Forum.

[7]  Jürgen Schmidhuber,et al.  Long Short-Term Memory , 1997, Neural Computation.

[8]  Wil M. P. van der Aalst,et al.  Process Mining - Discovery, Conformance and Enhancement of Business Processes , 2011 .

[9]  Alessandro Sperduti,et al.  Time and activity sequence prediction of business process instances , 2016, Computing.

[10]  Matthias Weidlich,et al.  Queue Mining - Predicting Delays in Service Processes , 2014, CAiSE.

[11]  Stefan Jablonski,et al.  Dynamic guidance enhancement in workflow management systems , 2012, SAC '12.

[12]  Jörg Becker,et al.  Designing and implementing a framework for event-based predictive modelling of business processes , 2014, EMISA.

[13]  Jürgen Schmidhuber,et al.  Framewise phoneme classification with bidirectional LSTM and other neural network architectures , 2005, Neural Networks.

[14]  Christian Sturm,et al.  Distributed Multi-Perspective Declare Discovery , 2017, BPM.

[15]  Stefan Jablonski,et al.  Towards Simulation- and Mining-based Translation of Resource-aware Process Models , 2016, Business Process Management Workshops.

[16]  Edward F. Watson,et al.  Business Process Automation , 2009, Handbook of Automation.

[17]  Fabrizio Maria Maggi,et al.  Predictive Monitoring of Business Processes , 2013, CAiSE.

[18]  P. Mahadevan,et al.  An overview , 2007, Journal of Biosciences.

[19]  Marlon Dumas,et al.  Predictive Business Process Monitoring with LSTM Neural Networks , 2016, CAiSE.

[20]  Guigang Zhang,et al.  Deep Learning , 2016, Int. J. Semantic Comput..

[21]  Heekuck Oh,et al.  Neural Networks for Pattern Recognition , 1993, Adv. Comput..

[22]  Nitish Srivastava,et al.  Dropout: a simple way to prevent neural networks from overfitting , 2014, J. Mach. Learn. Res..

[23]  Jan Mendling,et al.  Discovery of Multi-perspective Declarative Process Models , 2016, ICSOC.

[24]  Jörg Becker,et al.  Comprehensible Predictive Models for Business Processes , 2016, MIS Q..

[25]  Klaus Pohl,et al.  Comparing and Combining Predictive Business Process Monitoring Techniques , 2015, IEEE Transactions on Systems, Man, and Cybernetics: Systems.

[26]  Fabrizio Maria Maggi,et al.  Complex Symbolic Sequence Encodings for Predictive Monitoring of Business Processes , 2015, BPM.