基于多标记学习的中医药方功能预测研究

Study of Function Prediction for Traditional Chinese Medicine Prescriptions Based on Multi-label Learning

中医是西方医学之外,一门独立的医学体系。经过几千年的发展,中医领域已经积累了大量成熟的医药方剂,如何从这些中医药方中提取出有意义且潜在有用的知识,是目前机器学习及数据挖掘领域所关注的诸多热点应用问题之一。论文考虑采用多标记学习技术建立中医药方与药方功能间的映射关系,从而开发相对精确的中医药方功能预测模型。通过多种多标记学习算法在超过一千个成熟的中医药方上的建模结果显示:多标记学习技术有能力相对精确地根据中医药方预测其功能。

Different from the western medicine,the Traditional Chinese Medicine(TCM)has created an independent medi-cine system.During the past thousands years,TCM has accumulated a mass of mature prescriptions.Then,how to extract some po-tentially valuable knowledge from these TCM prescriptions has attracted many researchers in machine learning and data mining fields.This study considers to construct the mapping relationship between TCM prescription and its efficacies and to develop approxi-mately accurate TCM prescription efficacy prediction model by using multi-label learning techniques.The experiments are conduct-ed on multiple different multi-label learning algorithms and more than 1000 mature TCM prescriptions.The results show that the multi-label learning technique is capable of providing approximately accurate prediction for TCM prescriptions'efficacies.