监督主题模型的临床文本挖掘和药效预测
Clinical Text Mining and Efficacy Prediction Studies Using Supervised Topic Models
谢新平 1裴洋洋 1姜晓东 2王红强3
作者信息
- 1. 安徽建筑大学数理学院,安徽合肥 230601
- 2. 中国科学技术大学第一附属医院肿瘤科,安徽合肥 230031
- 3. 中国科学院合肥物质科学研究院智能所,安徽合肥 230031
- 折叠
摘要
患者的临床文本隐含着个体与药效之间的密切联系.针对临床上抗癌药效精准性问题,基于有监督隐含狄利克雷分布(Supervised Latent Dirichlet Allocation,SLDA)构建了一种用于药效二分类预测新方法——伯努利-监督隐含狄利克雷分布(Bernoulli-SLDA,B-SLDA),该模型获得患者临床文本的特征表示,学习到与对应药效标签的映射关系.实验结果表明,对比传统的特征提取方法,所提方法提高了抗肿瘤药物药效预测性能.
Abstract
The clinical texts of patients imply a close relationship between individuals and drug efficacy.In order to solve the problem of the accuracy of anticancer drug efficacy in clinical practice,based on the Supervised Latent Dirichlet Allocation(SLDA),a new method for pharmacodynamic dichotomous prediction B-SLDA was constructed,in which the characteristic representation of patients'clinical texts was obtained,and the mapping relationship with the corresponding pharmacodynamic labels was learned to achieve the prediction purpose.The experimental results show that compared with the traditional feature extraction methods,the proposed method improves the performance of anti-tumor drug efficacy prediction.
关键词
监督主题模型/监督隐含狄利克雷分布(SLDA)/药效预测/文本分类Key words
supervised topic model/Supervised Latent Dirichlet Allocation(SLDA)/efficacy prediction/text classification引用本文复制引用
出版年
2024
NETL
NSTL
万方数据