A Closer Look at Deep Learning Survival Prediction for High-Throughout Data

Research on survival prediction with deep learning has recently emerged. We found that deep learning has great advantages in the process of survival prediction. In this article, we compare traditional deep learning with two baseline models. They are Cox-nnet and DeepSurv. To summarize: When mining TCGA (The Cancer Genome Atlas) high-throughput genetic data sets for analysis: the results obtained by the simple and easy-to-use model are more credible, and its concordance index (evaluation of survival prediction model indicators) is also higher. We compare the loss curves of the two models and give our analysis.