Thursday, 16 February 2017, 14:00
IT Bldg (Akadeemia tee 15A), room 507AB
Abstract: Predictive business process monitoring is concerned with predicting future states or properties of ongoing executions of a business process, based on past executions thereof. Such predictions can range from predicting which activity will be performed next, when, and who will perform it, to predicting the remaining execution time or the final outcome of the process. For example, in an order-to-cash process, predictive monitoring techniques can be used to predict how likely is it that a purchase order will be fulfilled on time, or how likely is it that the customer will be satisfied after fulfillment of the order. In this talk, we will present a framework for conceptualizing and addressing predictive process monitoring problems using various machine learning techniques, ranging from classical classification techniques (e.g. random forests), to Hidden Markov Models and Recurrent Neural Networks. We will also present an empirical evaluation of the relative performance of these techniques and discuss their relative applicability and limitations.