A0542
Title: Deep neural network based accelerated failure time models with high-dimensional data
Authors: Gwangsu Kim - Jeonbuk National University (Korea, South) [presenting]
Abstract: An accelerated failure time (AFT) model assumes a log-linear relationship between failure times and a set of covariates. In contrast to other popular survival models that work on hazard functions, the effects of covariates are directly on failure times, whose interpretation is intuitive. Also, deep neural networks (DNN) have received focal attention over the past decades and have achieved remarkable success in various fields. DNNs have a number of notable advantages and have been shown to be particularly useful in addressing nonlinearity. By taking advantage of this, the method proposes to apply DNN in fitting AFT models using a Gehan-type loss combined with a sub-sampling technique. An extensive simulation study investigates the finite sample properties of the proposed DNN and rank-based AFT model (DeepR-AFT). DeepR-AFT shows superior and interesting performance, especially in high-dimensional data.
