<ul><li>Current approaches for suffix prediction in business processes focus on predicting a single, most likely suffix.</li><li>Proposed probabilistic suffix prediction offers a probability distribution of suffixes, addressing limitations of single predictions in uncertain or variable processes.</li><li>Uncertainty-Aware Encoder-Decoder LSTM (U-ED-LSTM) and Monte Carlo suffix sampling algorithm are key components of the proposed approach.</li><li>Evaluation of U-ED-LSTM shows good predictive performance and calibration on real-life event logs, highlighting the effectiveness of probabilistic suffix predictions in capturing uncertainties.</li></ul>

An Uncertainty-Aware ED-LSTM for Probabilistic Suffix Prediction

Discover more