Event suffix and remaining time prediction are sequence to sequence learning tasks . They have wide applications in different areas such as economics, digital health, business process management and IT infrastructure monitoring . Recent deeplearning-based works for such predictions are prone to potentially large errors because of closed-loop training (i.e., the next event is conditioned on the ground truth of previous events) We harness the power of adversariallearning techniques to boost prediction performance . The results show improvements up to four times compared to the state of the state-art in suffix and . remaining timeprediction of event sequences, specifically in the realm of business processexecutions. We also show that the obtained improvements of . adversarial . trainingare superior compared to standard training under the same experimental setup . under the . experimental setup. We consider four real-lifedatasets and three baselines in our experiments in our . experiments. The results showed improvements in adversarial training were superior compared . to standard . to the standard training .

Author(s) : Farbod Taymouri, Marcello La Rosa, Sarah M. Erfani

Links : PDF - Abstract

Code :

https://github.com/alsoj/Recommenders-movielens


Coursera

Keywords : training - event - improvements - standard - compared -

Leave a Reply

Your email address will not be published. Required fields are marked *