基于决策树集成学习在癌症风险分层中的应用
Application of Decision Tree-based Ensemble Learning in Cancer Risk Stratification
殷清燕 1车露美 1刘星宇2
作者信息
- 1. 西安建筑科技大学 西安 710055
- 2. 西安邮电大学 西安 710121
- 折叠
摘要
论文提出了基于决策树集成学习的癌症风险分层方法,并将其应用在TCGA癌症数据集.首先对原始数据进行数据清洗与预处理,构建基于决策树集成学习的生存分析模型;针对每个算法进行贝叶斯超参数优化,获得最优的C-in-dex和time-dependent AUC性能评价值,结果表明随机生存森林RSF和梯度提升树GBM性能较好.基于RSF和GBM风险评分进行癌症风险分层,在识别高危和低危患者群体方面具有显著作用.
Abstract
The paper proposes a cancer risk stratification method based on decision tree ensemble learning,and applies it on the TCGA cancer data set.Decision tree ensemble-based survival analysis model is constructed on the preprocessed data set,and the optimal hyperparameter by is chosen Bayesian optimization method.C-index and time-dependent AUC evaluation values show that random forest(RSF)and gradient boosting tree(GBM)are better than other algorithms.It shows that the cancer risk stratifica-tion method based on RSF and GBM risk scores plays a significant role in identifying high-risk and low-risk patients.
关键词
集成学习/随机生存森林/梯度提升树/生存分析/风险分层Key words
ensemble learning/random survival forest/gradient boosting tree/survival analysis/risk stratification引用本文复制引用
基金项目
西安市科技局项目(2019218114GXRC017CG018-GXYD17.6)
出版年
2024
国家科技期刊平台
NSTL
万方数据