[AFT-XGBoost] Survival regression with AFT model in XGBoost

Time: Thursday 17-Sept-2020 16:00 (This is a past event.)


Artifacts
slides: please to see content

Motivation / Abstract
In marketing modeling, time-to-event models occupy a very prominent role in deciding whom to target during a campaign and, just as importantly, when to target. In the paper serving as the foundation for this talk, the speakers (along with Dr. Toby Hocking) have developed a novel adaptation of an Accelerated Failure Time (AFT) model that integrates with XGBoost. This, in turn, paves the way for implementations of tree-based models that are able to offer support for survival regression. Furthermore, use of NVIDIA GPUs allows for a substantial speedup in training times. 
Stream Categories:
 ML in Marketing