关键词: 电子病历, 疾病危重度, 动态预测, 特征挖掘, 朴素贝叶斯, 数据挖掘

Abstract:

This article has built a dynamic forecast model suitable for all kinds of common serious diseases on the basis of the electronic medical record (EMR) big data. The dynamic forecast model was built according to the data mining methods, including nave Bayes theory, corrrelation analysis, and information gain. Based on the experiments in the database of EMR also known as MIMIC III, the effect of the dynamic forecast model was tested, and the main distinguishing characteristics of the severity of diease were screened. Data mining experiment results show that the 38 distinguishing characteristics discovered in this research is discernible to severity of dieases. Meanwhile, this dynamic forecast model is effective for forecasting the severity of various common severe dieases, and the short-term forecasting accuracy is especially high.

Key words: Electronic medical record, Severity of disease, Dynamic forecasting, Feature mining, Naive Bayes, Data mining