Skip to main content

Neural Conditional Event Time Models

Publication ,  Journal Article
Engelhard, M; Berchuck, S; D'Arcy, J; Henao, R
April 3, 2020

Event time models predict occurrence times of an event of interest based on known features. Recent work has demonstrated that neural networks achieve state-of-the-art event time predictions in a variety of settings. However, standard event time models suppose that the event occurs, eventually, in all cases. Consequently, no distinction is made between a) the probability of event occurrence, and b) the predicted time of occurrence. This distinction is critical when predicting medical diagnoses, equipment defects, social media posts, and other events that or may not occur, and for which the features affecting a) may be different from those affecting b). In this work, we develop a conditional event time model that distinguishes between these components, implement it as a neural network with a binary stochastic layer representing finite event occurrence, and show how it may be learned from right-censored event times via maximum likelihood estimation. Results demonstrate superior event occurrence and event time predictions on synthetic data, medical events (MIMIC-III), and social media posts (Reddit), comprising 21 total prediction tasks.

Duke Scholars

Publication Date

April 3, 2020
 

Citation

APA
Chicago
ICMJE
MLA
NLM
Engelhard, M., Berchuck, S., D’Arcy, J., & Henao, R. (2020). Neural Conditional Event Time Models.
Engelhard, Matthew, Samuel Berchuck, Joshua D’Arcy, and Ricardo Henao. “Neural Conditional Event Time Models,” April 3, 2020.
Engelhard M, Berchuck S, D’Arcy J, Henao R. Neural Conditional Event Time Models. 2020 Apr 3;
Engelhard, Matthew, et al. Neural Conditional Event Time Models. Apr. 2020.
Engelhard M, Berchuck S, D’Arcy J, Henao R. Neural Conditional Event Time Models. 2020 Apr 3;

Publication Date

April 3, 2020