计算机科学
分类器(UML)
图形
人工智能
多层感知器
鉴定(生物学)
药方
机器学习
数据挖掘
人工神经网络
模式识别(心理学)
医学
理论计算机科学
植物
药理学
生物
作者
Amin Fu,Jishun Ma,Chuansheng Wang,Chunyan Zhou,Zuoyong Li,Shenghua Teng
标识
DOI:10.1007/978-3-031-20096-0_1
摘要
The Traditional Chinese Medicine Health Status Identification plays an important role in TCM diagnosis and prescription recommendation. In this paper, we propose a method of Status Identification via Graph Attention Network, named SIGAT, which captures the complex medical correlation in the symptom-syndrome graph. More specifically, we construct a symptom-syndrome graph in that symptoms are taken as nodes and the edges are connected by syndromes. And we realize automatic induction of symptom to state element classification by using the attention mechanism and perceptron classifier. Finally, we conduct experiments by using hamming loss, coverage, 0/1 error, ranking loss, average precision, macro-F1 score, and micro-F1 score as evaluation metrics. The results demonstrate that the SIGAT model outperforms comparison algorithms on Traditional Chinese Medicine Prescription Dictionary dataset. The case study results suggest that the proposed method is a valuable way to identify the state element. The application of the graph attention network classification algorithm in TCM health status identification is of high precision and methodological feasibility.
科研通智能强力驱动
Strongly Powered by AbleSci AI