Study On Evaluation Of Evidence-based Data Mining Method Of Traditional Chinese Medicinal Property And Efficacy

Research Objectives:Through the collection and collation of domestic data mining methods in the study of traditional Chinese medicinal property and efficacy research,the evidence-based method was used to statistically analyze the literature on property and efficacy,and the comparison results of traditional Chinese medicinal property and efficacy data mining methods were visualized and objective.In order to provide reference and inspiration for the application of data mining methods in information mining of traditional Chinese medicine.Research methods:Search CNKI,VIP,and WanfangData,search time range from Database establishment to September 9,2017,search terms Different expressions of 52 kinds of data mining methods such as principal component analysis,factor analysis,Bayesian network,etc.,as well as Chinese medicine efficacy,Chinese medicine medicinal properties,Chinese medicine efficacy,the search language is limited to Chinese,and the retrieved documents are imported into NoteExpress software.The literature database standardizes the data mining names,and summarizes the basic characteristics,application trend rules,application content distribution and applied data mining software and basic functions of the included documents through mathematical statistics methods.The literature is screened according to the pre-set exclusion criteria,and the basic data and important information are extracted according to the designed data extraction form.The literature screening and data extraction are performed independently by two people,and the disputed data is sought for the three parties assisted in the decision.The Excel 2016 and Stata 15.0 software were used to analyze and systematically evaluate the extracted data,and the odd ratio(OR)value was used as the effect index to evaluate the application effect of the data mining method in the identification of cold medicine of traditional Chinese medicine.Sensitivity analysis was performed using a random effects model and a fixed effect model,respectively.Through the mathematical statistics method,the data mining method is used to analyze the application effect of the single-flavor Chinese medicine efficacy prediction and the Chinese medicine compound efficacy prediction,to clarify the research type and the research object,and to filter the literature according to the set inclusion criteria and exclusion criteria,and according to the The designed data extraction form extracts information,and the literature screening and data extraction are performed independently by two people.For the disputed data,third party assistance judgment is sought.Statistical analysis was carried out on the data mining methods,modeling basis,and forecasting efficiency of related documents.Research results:The most popular applications of data mining methods in traditional Chinese medicinal property and efficacy mining are,cluster analysis,principal component analysis,etc.,and the number of documents published in the field of Chinese medicinal property and efficacy research from 2000 to 2018 is increasing year by year.In particular,after 2015,the growth rate is particularly fast,the quality of the literature is constantly improving,and the aspects involved are becoming more extensive.In terms of the identification of cold and hot drug properties of traditional Chinese medicine,the meta-analysis,the comparison of the prediction effects of “Support Vector Machine” and “Logistic-Discrimina Analysis”,the total prediction accuracy rate(OR=1.74,95% CI:(1.13~2.68);“Support Vector Machine” and The " Principal Component Analysis-Linear-Discrimina Analysis " was compared and the OR values were 1.47 > 1,95% CI(0.95,2.25),P = 0.081 > 0.05."Support Vector Machine" and " Partial Least Squares-Discrimina Analysis " are combined and the OR values are 1.15>1,95% CI(0.75,1.77),P=0.512>0.05);“Logistic-Discrimina Analysis” and “Principal Component Analysis-Linear-Discrimina Analysis" comparison combination and OR value is 1.19>1,95% CI(0.77,1.83),P=0.440>0.05;" Logistic-Discrimina Analysis " and " Partial Least Squares-Discrimina Analysis" combined OR value is 1.51>1,95% CI(0.98~2.32),P=0.063>0.05;"Principal Component Analysis-Linear-Discrimina Analysis " and " Partial Least Squares-Discrimina Analysis " combined OR value of 1.27>1,95% CI(0.83,1.95),P= 0.275>0.05,the remaining two groups abandoned the synthesis due to the heterogeneity.The data mining methods commonly used in data mining for power prediction are neural networks and Bayesian networks.It has been found that neural networks have been used in the prediction of efficacy and compound efficacy in single-flavor Chinese medicine.Bayesian networks have only been applied to single-flavors.Chinese medicine efficacy prediction.The commonly used modeling basis for efficacy prediction is usually pharmacological indicators,menstruation,efficacy,indications,doses,etc.The prediction rates of the models built in the current literature reports are all high.Research Conclusion:At present,there are many data mining methods in the application of traditional Chinese medicinal property and efficacy research,the prediction is accurate and high,to a certain extent,the data mining method is more applicable in the field of traditional Chinese medicinal property and efficacy research.At present,the research has been mainly focused on the identification of cold and hot medicine.After comparison and analysis,the application results of Support Vector Machine in the application of cold and hot drug identification are more prominent.The results of other statistical data mining methods in this aspect are quite comparable.There is no obvious difference.At present,there are relatively few studies on the prediction of efficacy of traditional Chinese medicines.The main application methods are Bayesian networks and Neural networks,and the prediction accuracy is high.The Neural network is applicable to the prediction of the efficacy of single Chinese medicine and the efficacy prediction of traditional Chinese medicine.It has become the primary data mining method in current power prediction research.Because of the insufficient number of such studies,comparisons cannot be made,and as evidence increases,research will present more reliable comparisons.