隐结构模型在慢性胃炎辨证中的应用探索——基于EM算法的因子分析方法

Explore the Application of Latent Structure in Chronic Gastritis --Based on EM Algorithm and Factor Analysis

名老中医在长年的临床实践中积累了大量的宝贵经验,而这些经验都隐含在众多的临床病历中,机器学习是挖掘出这些隐含经验非常有效的工具,因此利用机器学习技术挖掘出隐含在大量病历资料中的临床经验,对于名老中医经验传承具有非常重要的价值。隐结构模型是张连文教授提出的一种模型,它能够较好地符合中医辨证理论。本文在其方法上进行了一定的简化和改进,并应用于慢性胃炎辨证。主要是采用基于EM（expectation maximum,最大期望）算法的因子分析方法处理病案数据,从而得到慢性胃炎辨证的隐结构,提高了学习速度和模型的准确性。

Famous herbalist doctors accumulate a lot of precious experience during long-period clinical diagnosis. Normally, this kind of experience is hidden in a great deal of clinical medical records. The machine learning technique is a very effective tool to mine such experience. And mining clinical experience hidden in medical records by machine learning technique is of the important value for inheriting experience from famous herbalist doctors. Latent structure model was proposed by Professor Zhang Lianwen, which accords with the syndrome differentiation of traditional Chinese medicine （TCM）. In this paper, the latent structure model was simplified and improved, and applied to the syndrome differentiation of chronic gastritis. Factor analysis method based on the EM algorithm was adopted to analyze the data for the medical records, accordingly the latent structure of syndrome differentiation of the chronic gastritis was obtained, which improved the learning speed and accuracy of model.