Dealing with censorship in neural network models

摘要: Feedforward neural networks have recently been considered as nonlinear tools for modelling survival data. This requires handling censorship, since it is inherent in many such studies, including following breast cancer surgery which the subject of this study. Previous studies with concentrated on beyond a single time interval, or extended Cox regression model separate output nodes each often ignoring censorship altogether. There evidence to suggest that censored data can introduce significant bias into estimates survival. We report using traditional MLP architecture, where issue has removed from network structure and dealt part structure. also show two coding methods missing are tested investigate their effect accuracy estimates.