基于LDA模型的疾病特征识别方法

Disease feature identification approach based on LDA model

目前基于病历的疾病特征提取方法有很多,但是这些方法存在病历数据离散化程度高、描述语言标准不统一、疾病特征的提取困难等问题。鉴于此,提出一种基于LDA模型的疾病特征识别方法。该方法能同时建模病历、疾病、特征三者之间的关系,得出病历-疾病和疾病-特征两个分布矩阵,从而达到疾病特征识别的目的。实验表明该方法的疾病特征识别准确率高于ID3算法和C4.5算法,达到了良好的疾病特征识别效果。

There are many ways of extracting disease characteristics based on disease records, but these ways have many problems such as high discrete of disease record data, non - unified description language and difficulty in extracting disease features, etc.. How to i- dentify disease features from electronic medical records to assist doctors in diagnosing diseases is an important research in the field of computer - aided diagnosis. This paper proposes a LDA model - based disease feature recognition method, which can model medical re- cords, diseases and characteristics at the same time, and obtain the two distributions of disease record - disease and disease - feature so as to achieve the goal of disease feature recognition. Experiments show that the method has higher recognition accuracy than ID3 algo- rithm and C4.5 algorithm, and achieves good feature recognition results.