[1]李文琦,黄水平#,李海朋.BP神经网络和Cox比例风险模型在生存分析应用中的比较*[J].郑州大学学报(医学版),2014,(06):822.
 LI Wenqi,HUANG Shuiping,LI Haipeng.Comparison between BP neural network and Cox proportional hazard model in survival analysis[J].JOURNAL OF ZHENGZHOU UNIVERSITY(MEDICAL SCIENCES),2014,(06):822.
点击复制

BP神经网络和Cox比例风险模型在生存分析应用中的比较*
分享到:

《郑州大学学报(医学版)》[ISSN:1671-6825/CN:41-1340/R]

卷:
期数:
2014年06期
页码:
822
栏目:
应用研究
出版日期:
2014-11-20

文章信息/Info

Title:
Comparison between BP neural network and Cox proportional hazard model in survival analysis
作者:
李文琦黄水平#李海朋
徐州医学院公共卫生学院流行病与卫生统计学教研室 徐州 221002
Author(s):
LI Wenqi HUANG Shuiping LI Haipeng
Department of Epidemiology and Medical Statistics, School of Public Health, Xuzhou Medical College, Xuzhou 221002
关键词:
BP神经网络Monte Carlo 模拟生存分析预测模型Cox比例风险模型
Keywords:
BP neural network Monte Carlo simulation survival analysis forecast model Cox proportional hazard model
分类号:
R195.1
摘要:
目的:比较BP神经网络模型和Cox比例风险模型在生存分析中的预测性能,进一步探讨BP神经网络模型在生存分析中的应用。方法:采用Monte Carlo模拟数据集,如不同样本量、不同删失比例、不同协变量间关系及是否满足等比例风险假定的理论研究和胃癌根治术患者预后预测的实例分析,分别建立BP神经网络模型和Cox比例风险模型,最终使用一致性指数C对其预测性能进行比较。结果:当样本量为100、删失比例为60%、80%及样本量为300、删失比例为80%时,BP神经网络模型的预测性能高于Cox比例风险模型(P<0.05)。协变量不满足等比例风险假定、协变量间存在三维交互作用和非线性关系时,BP神经网络模型的预测性能较Cox比例风险模型好(P<0.05)。实例研究中发现,BP神经网络模型预测的一致性指数C(0.835)高于Cox比例风险模型(t配对=4.311,P<0.001)。结论:BP神经网络模型在生存分析的应用中对样本删失比例、是否满足PH假定、协变量间复杂交互作用和非线性关系具有非特异性,对资料限制较少,且预测一致性高,值得在生存分析中进一步推广应用。
Abstract:
Aim: To compare their prediction performance of BP neural network model and Cox proportion hazard model in survival analysis and to explore the superiority of BP neural network model in survival analysis. Methods: Monte Carlo was used to generate the data sets under the condition of different sample size, different degree of censoring, number of variable and interactions, nonlinear effect, distinct distribution of covariate and proportional vs nonproportional hazard. Then BP neural network model and Cox model were built, and their prediction performance was compared using concordance index C. Results: In the research on simulation data sets, when the sample size of 100, proportion of censoring of 60%, 80%, and sample size of 300, proportion of censoring of 80%, BP neural network model performed superior to Cox model(P<0.05). And when the covariates don′t meet PH assumption and had threeway interaction, nonlinear effect, BP neural network performed superior to Cox model(P<0.05). In the real data, BP neural network model′s concordance index was 0.835, which performed superior to Cox model(tpaired=4.311,P<0.001). Conclusion: For the small sample size, high and the covariates don′t meet PH assumption and has threeway interaction, nonlinear effect data sets, BP neural network has better advantage than Cox model. It is worth to popularize further in survival analysis.

参考文献/References:

[1]王娟.截尾分位数回归模型及其在生存分析中的应用[D].太原:山西医科大学,2009. [2]高蔚,施侣元.人工神经网络流行病学应用进展[J].中华预防医学杂志,2000,34(6):373 [3]Cox DR,Oakes D. Analysis of survival data[M]. London:Chapman and Hall,1984:201 [4]钱俊.生存分析中删失数据比例对Cox回归模型影响的研究[D].广州:南方医科大学,2009:95

备注/Memo

备注/Memo:
*江苏省科技厅资助项目BE2011647;#通讯作者,女,1963年10月生,硕士,教授,研究方向:统计方法在流行病学研究中的应用,Email:hsp@xzmc.edu.cn
更新日期/Last Update: 2014-11-26