摘要
针对当前重症患者预后相关因素的研究主要集中于线性回归分析,构建了一基于贝叶斯网络的老年重症患者预后评估系统。提出了一种基于最小描述长度与K2算法的贝叶斯方法,以获得较优的网络结构;并利用最大似然估计进行参数学习。四折交叉抽样的实验结果表明,所构建系统的预测精度比传统的BP神经网络和基于K2的贝叶斯网络学习分别提高了6.87%和27.20%.这将为医生预测高龄患者在ICU治疗中的受益程度提供临床决策支持。
In critically ill patients, the elderly accounts for a large proportion and takes up more ICU resources, but the treatment effect and prognosis are not clear. Therefore, prognosis research for elderly critical patients is important. At present, most research focuses on the prognostic factors by using the regression analysis which often assumes a linear relationship between death and various risk factors for the sake of simplifying problems. The Bayesian network is an effective tool for uncertain reasoning and nonlinear analysis, and generated models are comprehensible. In this paper, an evaluation model of the prognosis for elderly critical patients based on Bayesian network was constructed. A Bayesian approach based on Minimum Description Length （MDL） and K2 algorithm was proposed to obtain the optimal network structure, and then the maximum likelihood method was used for parameter learning. At last, Bayesian inference was employed to achieve the prediction results. Four-fold cross sampling experiment results show that the prediction accuracy of the model presented in this paper was superior to both conventional BP Neural Networks and K2 algorithm based on Bayesian learning method, and the prediction accuracy has been improved by 6.87% and 27.20%, respectively. It is helpful for the doctors to estimate how much elderly critical patients would benefit from ICU treatment.
出处
《太原理工大学学报》
CAS
北大核心
2014年第3期352-357,共6页
Journal of Taiyuan University of Technology
基金
国家自然科学基金资助项目(61300107)
广东省自然科学基金资助项目(S2012010010212)
关键词
贝叶斯网络
参数学习
预后评估
Bayesian networks
estimation parameter learning
evaluation of the prognosis
