Adapting machine learning techniques to censored time-to-event health record data: A general-purpose approach using inverse probability of censoring weighting
详细信息 查看全文
- 作者：David M. Vocka ; vock@umn.edu" class="auth_mail" title="E-mail the corresponding author ; Julian Wolfsona ; julianw@umn.edu" class="auth_mail" title="E-mail the corresponding author ; Sunayan Bandyopadhyayb ; band0064@umn.edu" class="auth_mail" title="E-mail the corresponding author ; Gediminas Adomaviciusc ; gedas@umn.edu" class="auth_mail" title="E-mail the corresponding author ; Paul E. Johnsonc ; johns021@umn.edu" class="auth_mail" title="E-mail the corresponding author ; Gabriela Vazquez-Benitezd ; gabriela.x.vazquezbenitez@healthpartners.com" class="auth_mail" title="E-mail the corresponding author ; Patrick J. O&rsquo ; Connord ; patrick.j.oconnor@healthpartners.com" class="auth_mail" title="E-mail the corresponding author
- 关键词：Censored data ; Electronic health data ; Inverse probability weighting ; Machine learning ; Risk prediction ; Survival analysis
- 刊名：Journal of Biomedical Informatics
- 出版年：2016
- 出版时间：June 2016
- 年：2016
- 卷：61
- 期：Complete
- 页码：119-131
- 全文大小：538 K
文摘
- •
Right-censored outcomes are common in biomedical prediction problems.
- •
We discuss adapting machine learning (ML) algorithms to these outcomes using IPCW.
- •
IPCW is a general-purpose approach which can be applied to many ML techniques.
- •
ML with IPCW leads to more accurate predictive probabilities than ad hoc approaches.