The interest in Machine Learning applications in Process Mining has seen increasing growth in the last few years along with the relevance of the ICPM conference. Thus, this workshop offers a focused environment to discuss new approaches, applications and their results to a wide audience, composed of researchers and practitioners. ML4PM will be held in Padua, in conjunction with the ICPM conference.
This workshop invites papers that present works that lay in the intersection between machine learning and process mining. The event provides a suitable environment to discuss new approaches presented by researchers and practitioners. Main themes include automated process modeling, predictive process mining, application of deep learning techniques and online process mining. The workshop will count with leading researchers, engineers and scientists who are actively working on these topics.
Topics of interest for submission include, but are not limited to:
Due to the exceptional circumstances of the COVID-19 outbreak, ICPM 2020 will be a fully virtual conference, with no travel involved. However, the entire program, including the co-located events, will be retained, and will not change. With the spirit of keeping the entire conference as interactive as possible, presentations will be given live using webinars. The presentations will also be broadcasted, and also available after the conference for off-line viewing. Attendees will be able to ask questions, which will be answered at the end of each presentation. When multiple sessions run in parallel (e.g., workshops), the conference will feature parallel virtual rooms.
Contributions to all calls should be submitted electronically to the Workshop management system connecting to ICPM EasyChair portal. At least one author of each accepted paper is expected to participate in the conference and present his/her work.
Submissions must be original contributions that have not been published previously. Authors are requested to prepare submissions according to the format of the Lecture Notes in Business Information Processing (LNBIP) series by Springer (http://www.springer.com/computer/lncs?SGWID=0-164-6-791344-0). Submissions must be in English and must not exceed 12 pages (including figures, bibliography and appendices). Each paper should contain a short abstract, clarifying the relation of the paper with the workshop topics, clearly state the problem being addressed, the goal of the work, the results achieved, and the relation to the literature.
Registrations are managed by the ICPM system
| Milestone | Deadline |
|---|---|
| Abstract Submission | September 1 2020 |
| Paper Submission | September 4 2020 |
| Notification of Acceptance | September 22 2020 |
| Submission of Camera Ready Papers | September 29 2020 |
| Workshop | October 5 2020 |
The availability of powerful techniques for supervised machine learning has boosted interest in their application to business process analysis, formulating BPM problems in terms of prediction and classification. While this approach has proven fruitful in some cases, several issues remain. The talk focuses on two major ones: variable memory and changing semantics. The first issue regards long and medium-term meta-learning effects, e.g., human or organizational learning that may invalidate classification inferences based on examples collected at a different time and are difficult to handle with classic model tuning. The second issue relates to the subsymbolic nature of supervised machine learning. Subsymbolic inference is fundamental nonmonotonic, it does not admit a precise formal conceptual description and it does not allow to transfer conclusions from one case to another. For these reasons, the opportunities made available can be fully realized by clearly understanding the pitfalls we have to face.
| Time | Presentation Title |
|---|---|
| 12:00 | Ernesto Damiani's Keynote: Applying AI to BPM: opportunities and pitfalls |
| 13:00 | BREAK |
| 13:30 | Predicting Remaining Cycle Time from Ongoing Cases: A Survival Analysis-based Approach |
| 13:45 | Time Matters:Time-Aware LSTMs for Predictive Business Process Monitoring |
| 14:00 | A preliminary study on the application of Reinforcement Learning for Predictive Process Monitoring |
| 14:15 | An Alignment Cost-Based Classification of Log Traces Using Machine-Learning |
| 14:30 | BREAK |
| 14:45 | Improving the Extraction of Process Annotations from Text with Inter-Sentence Analysis |
| 15:00 | Case2vec: Advances in Representation Learning for Business Processes |
| 15:15 | Supervised Conformance Checking using Recurrent Neural Network Classifiers |
María Teresa Gómez, University of Seville
Rafael Accorsi, PwC Digital Services
Niek Tax, Booking.com
Josep Carmona, Universitat Politècnica de Catalunya
Ernesto Damiani, Khalifa University of Science Technology
Chiara Di Francescomarino, Fondazione Bruno Kessler
Antonella Guzzo, Università della Calabria
Mariangela Lazoi, University of Salento
Matthias Ehrendorfer, University of Vienna
Fabrizio Maria Maggi, University of Tartu
Paola Mello, Università di Bologna
Gabriel Marques Tavares, Università degli Studi di Milano
Emerson Cabrera Paraiso, Pontifícia Universidade Católica do Paraná
Bruno Bogaz Zarpelão, State University of Londrina
Irene Teinemaa, Booking.com
Michelangelo Ceci, University of Bari Aldo Moro
Natalia Sidorova, Eindhoven University of Technology
Domenico Potena, Università Politechnica delle Marche
