A novel lifelong machine learning-based method to eliminate calibration drift in clinical prediction models
作者：
Highlights：
• A study of clinical prediction model performance found changes in data distributions cause model performances to drift.
• A model updating method based on lifelong machine learning to solve the calibration drift caused by data drift was proposed.
• The effectiveness of the proposed method is verified on four datasets.
摘要
•A study of clinical prediction model performance found changes in data distributions cause model performances to drift.•A model updating method based on lifelong machine learning to solve the calibration drift caused by data drift was proposed.•The effectiveness of the proposed method is verified on four datasets.
论文关键词：Clinical prediction models,Lifelong machine learning,Knowledge distillation,Calibration,Cancer
论文评审过程：Received 18 April 2021, Revised 14 January 2022, Accepted 9 February 2022, Available online 12 February 2022, Version of Record 16 February 2022.
论文官网地址：https://doi.org/10.1016/j.artmed.2022.102256
