Open AccessPrediction of Length of Stay on the Intensive Care Unit Based on Bayesian Neural Network
Author(s) -
Jiansheng Fang,
Jiang Zhu,
Xiaoqing Zhang
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1631/1/012089
Subject(s) - interpretability , overfitting , artificial neural network , computer science , bayesian probability , machine learning , intensive care unit , artificial intelligence , generalization , data mining , medicine , mathematics , intensive care medicine , mathematical analysis
Predicting length of stay (LoS) accurately in the intensive care unit (ICU) is important to improve care quality and resource utilization. However, for LoS prediction, existing methods are facing main challenges, including uncertain prediction, generalization, interpretability, etc. In this paper, we utilize Bayesian neural network (BNN) to alleviate the above main challenges. The BNN introduces prior knowledge on the weights of neural networks and estimates outcomes from the predictive distribution after calculating the posterior distribution of weights. Extensive experiments on the eICU collaborative research database (eICU-CRD) show that the proposed method is competitive and more capable of anti-overfitting.