基于循证的中药药性判别数据挖掘方法评价

Evaluation of evidence-based traditional Chinese medicinal discriminant data mining method

目的:通过循证方法分析评价数据挖掘方法在中药性效中的应用效果。方法:通过检索万方、知网、维普数据库中的相关文献,按研究需求对文献进行筛选,并使用STATA 15.0对提取的资料进行分析。结果:共搜集文献2 064篇,最终纳入有效文献3篇,含6种数据挖掘方法,共进行8组比较。其中"支持向量机"和"回归判别分析"预测效果比较,总预测准确率为(OR=1.74,95%CI [1.13,2.68]);其中两组比较存在异质性;剩余5组无统计学意义。结论:"支持向量机"在中药药性判别中的预测效果优于"回归判别分析";其他比较组的方法的比较结果均无明显优势,需更多高质量的研究证实。

Objective: To analyze the application effect of data mining methods in the efficacy of traditional Chinese medicine through evidence-based methods. Methods: By searching relevant literatures in Wanfang, CNKI and VIP databases,the literature was screened according to research needs, and the extracted data were analyzed by STATA 15.0. Results: A total of2 064 articles were collected, and 3 valid articles were finally included, including 6 data mining methods. A total of 8 groups were compared. Among them, ‘SVM’ and ‘Logistic-DA’ predictive effect comparison, the total prediction accuracy rate(OR=1.74,95%CI[1.13, 2.68]);two groups are heterogeneous;the rest 5 groups are not statistical significance. Conclusion: The prediction effect of ‘SVM’ in the judgment of traditional Chinese medicine is better than ‘Logistic-DA’. There is no obvious advantage in the mothed of other comparative groups, and more high-quality studies are needed to confirm it.