Estimating Remaining Time of Business Processes with Structural Attributes of the Traces

Aburomman, Ahmad Lama, Manuel Bugarín-Diz, Alberto In this paper, we deal with one of the challenges in process mining enhancement: prediction of remaining times in a business process, which is a critical task for many organisations. Our approach consists of (i) defining a number of attributes on the business logs that capture structural information from the traces, (ii) extending the well-known annotated transition system model to annotate its states with the attributes values and (iii) applying linear regression for predicting the remaining time of the process for each state using the attributes values. Experiments with ten well-known real-life datasets show that our approach outperforms the baseline model defined in Van der Aalst et al. (Inf Syst 36(2):450–475, 2011) in the three metrics considered.