Bayesian Artificial Neural Networks for Survival Modeling
Artificial neural networks have been widely used in the field of pattern recognition within the past two decades.
Bayesian learning of artificial neural networks is very useful in overcoming many of the inherited problems in neural
networks, including the network over fitting, which is critical in obtaining generalized predictions with higher prediction
accuracies. Despite its importance, very few studies have used a Bayesian neural network for survival predictions. Accurate
prediction of patients survival is key for identifying the relevant treatment protocols in the various onset of cancers. In this
study, we demonstrate the use of Bayesian artificial neural networks for accurate survival predictions. In fact, we discuss
how to develop a piecewise constant hazard model using Bayesian neural networks. The uncertainties of the predictions are
captured using the error bars.
Keywords - Artificial Neural Networks, Bayesian Learning, Piecewise Constant Hazard Model, Survival Prediction,