Gaël Bernard, PhD student in the Department of Information Systems and Periklis Andritsos (former professor of Big Data & Analytics at HEC Lausanne, now at University of Toronto) won the best paper award at ADBIS 2019.
In their day-to-day operations, companies often have to deal with complex business processes. For instance, the customer service process might be composed of many activities and involve many stakeholders. What if, after observing few events we could predict how the process will most likely end? One could use such prediction to: (1) better inform customers about the next steps, (2) better prioritize business cases, (3) anticipate complaints and proactively take actions to reduce them. State-of-the-art approaches that rely on Machine Learning are opaque: they can predict which activities will occur, however, they cannot explain the rationale behind those predictions. The algorithm that Gaël and Periklis propose addresses this gap - it not only predicts which activities will happen, but it also delivers the business process model(s) used in getting the predictions. The business process model can then be used for interpretation purposes. This contribution is essential for the decision making processes in companies, since business analysts can be reluctant in trusting predictions they do not understand.