RT Dissertation/Thesis
T1 Deep Learning Models for Predictive Monitoring of Business Processes
A1 Rama Maneiro, Efrén
K1 Process mining
K1 predictive monitoring
K1 deep learning
K1 graph neural networks
K1 reinforcement learning
AB In this thesis, we enhance predictive monitoring in processmining through the use of advanced deep-learning techniques.By integrating Graph Neural Networks with Recurrent NeuralNetworks, we learn directly from the process model whilealso considering event sequences. We introduce two neuralmodels: the first aims to predict the next activity in a businessprocess, while the second forecasts the remaining sequence ofactivities until the case finishes. For the latter problem, anew Reinforcement Learning model is also proposed todynamically learn optimal activity selection strategies duringtraining. All models are rigorously validated using real-worldevent logs under a novel evaluation methodology to facilitaterobust and fair comparisons between different predictivemonitoring approaches.
YR 2023
FD 2023
LK http://hdl.handle.net/10347/32387
UL http://hdl.handle.net/10347/32387
LA eng
DS Minerva
RD 24 abr 2026