基于gSpan改进算法的中医辨证论治模式挖掘研究

Study on Pattern Mining of TCM Syndrome Differentiation and Treatment Based on Improved gSpan Algorithm

目的扩展经典的频繁子图挖掘算法以获得在中医学科中表现更好的数据挖掘效果,从而得出隐含在中医病案中的辨证论治模式。方法结合中医病案数据特征,扩展经典的图挖掘算法,对多个症状属性分别设置最小支持度阈值参数,再用扩展后的基于多重最小支持度的数据挖掘算法对数据集中蕴含的辨证论治模式进行挖掘。结果对3319条慢性阻塞性肺疾病(急性加重期)真实病案数据应用扩展的频繁子图挖掘算法,得到一系列该病相关的八纲辨证模式。与经典算法相比,扩展算法挖掘得到的辨证模式在模式维度和数量方面均明显提升。结论扩展后的频繁子图挖掘算法能够运用于中医辨证论治模式的挖掘,发现病案中隐含的辨证规律,且在模式完备性上具有比原始算法更好的效果。

Objective To extend the classic frequent subgraph mining algorithm to obtain a data mining method that performs better in TCM;To obtain the patterns of TCM syndrome differentiation and treatment implicit in the TCM medical records.Methods Combining with the characteristics of TCM medical records data and extending the classic frequent subgraph mining algorithm,data mining algorithm which set different minimum support threshold parameters for different symptom attributes was used to discover the patterns of TCM syndrome differentiation and treatment contained in the data set.Results The extended frequent subgraph mining algorithm was applied to the 3319 real medical records of chronic obstructive pulmonary disease(acute exacerbation period),and a series of patterns of syndrome differentiation of eight principles related to the disease were obtained.Compared with the classic algorithm,the patterns of TCM syndrome differentiation obtained by the extended algorithm had a significant improvement in the dimension and quantity of patterns.Conclusion The expanded frequent subgraph mining algorithm can be used in the TCM syndrome differentiation and treatment pattern mining as well as find the implicit syndrome differentiation rules in medical records,and it has a better effect than the original algorithm in the completeness of the patterns.