CEHMR: Curriculum learning enhanced hierarchical multi-label classification for medication recommendation

Mengxuan Sun, Jinghao Niu, Xuebing Yang, Yifan Gu, Wensheng Zhang

Abstract: Highlights•This paper formulates the medication recommendation task as a hierarchical multi-label classification problem.•The proposed CEHMR models the medication dependency by enabling classifiers to learn the prior hierarchy.•CEHMR measures the difficulty of each training example and progressively achieves a smoother training process.•The experimental testing performed on the MIMIC-III data set confirms the advantages of CEHMR.•Extensive experiments demonstrate its potential in dealing with the diversity of training example difficulties.