| Objective: To obtain the medication rules of traditional Chinese medicine in the treatment of depression syndrome by data mining related algorithms,and classify the depression syndrome types,in order to provide certain reference for the subsequent clinical treatment of depression syndrome.Methods:The medical cases related to depression syndrome were collected from the ancient and modern medical case cloud platform,China National Knowledge Infrastructure,Wanfang and VIP database.The search formula is: KY = ’depression syndrome’ OR KY = ’visceral mania’ OR KY = ’Globus Hysteriocus’ OR KY =’depression’ OR KY = ’anxiety disorder’ OR KY = ’hysteria’ OR KY = ’insomnia’ OR KY = ’depression’,search the database,The search period was from January 1,1990 to May 31,2022,and the search date was June 16,2022.The prescription data in medical records were extracted,and 877 prescription data that met the requirements were sorted into Excel to establish a prescription drug data sheet,and standardized treatment of traditional Chinese medicine was carried out.Microsoft Excel 2016,IBM SPSS Statistics 22.0 and IBM SPSS Modeler 18.0 were used to conduct data mining and statistical analysis of high-frequency drugs in the treatment of depression with traditional Chinese medicine.The information of the four diagnoses of depression syndrome in medical records was obtained,and the symptoms of 1010 medical cases that met the requirements were standardized,and the symptom information was coded and represented by "0" or "1".Finally,the random forest and artificial neural network classification models were constructed in Jupyter notebook by Python.A confusion matrix was used to evaluate the model.Results:The top 5 Chinese herbs in frequency were Radix bupleurum,Radix paeoniae alba,Poria cocos,Angelica sinensis and Pinellia tuber,with a total frequency of 1900,accounting for 29.15% of the total frequency of high-frequency drugs.The frequency of high frequency drugs from high to low was: tonifying deficiency drugs,regulating qi drugs,clearing heat drugs,calming drugs,activating blood circulation and removing blood stasis drugs,promoting water and dampness drugs,eliminating phlegm,relieving cough and wheezing drugs,relieving surface drugs,soothing liver and reducing wind,opening orifices and eliminating dampness drugs.Among them,tonic deficiency drugs and qi regulating drugs were the most common,accounting for42.57% of the total frequency of high frequency drugs.The analysis of high frequency drug properties for depression syndrome showed that cold,warm and flat were the top three drugs from high to low,and cold accounted for 41.47% of the total high frequency drugs.The taste of the medicine was bitter,spicy,sweet,sour,light,astringent and salty.The homing meridians were spleen meridian,liver meridian,heart meridian,lung meridian,gallbladder meridian,stomach meridian,kidney meridian,Sanjiao meridian,large intestine meridian,pericardium meridian,small intestine meridian and bladder meridian in turn.The results of 2,5,7,11 cluster analysis and second and third order association rules were also obtained.In the construction of the classification model of depression syndrome types,there are two main modules.The first module is the statistical analysis of the basic information of depression syndrome types,mainly including the distribution of depression syndrome types,gender,age,symptoms,pulse signs,tongue signs and fur signs.The second module was the construction of depression classification model.Python was used to construct random forest and artificial neural network classification models in Jupyter notebook,and the corresponding confusion matrix diagram was drawn to obtain the overall accuracy,precision,recall and F1 value of the corresponding syndrome.By comparing the two classification models,it was found that the overall accuracy of the classification model was high,and the obtained values were between 89% and 97%.The overall accuracy of the syndrome classification model constructed by the random forest algorithm was 89.44%,in which the stagnation of liver qi was 95%,the stagnation of qi was 82.05%,the stagnation of spitto-qi was 89.29%,and the deficiency of mind and mind was 85.07%.Heart and spleen deficiency 89.74%,heart and kidney Yin deficiency 95.16%;The overall accuracy of syndrome classification model constructed by artificial neural network algorithm was 96.03%,including100% stagnation of liver qi,92.31% stagnation of qi and fire,96.43% stagnation of phlegm and qi,91.04% deficiency of mind and spirit,97.44% deficiency of heart and spleen,100% deficiency of heart and kidney Yin.Conclusion: The overall medication was mainly for tonifying deficiency and regulating qi,with bitter,spicy and sweet flavors,both warm and cold,mostly cold,and mostly targeting the spleen,liver and heart meridians.The treatment mainly included invigorating the spleen,soothing the liver and regulating qi,drying dampness and eliminating phlegm,clearing heat and reducing fire,nourishing the heart and calming the mind.Cluster analysis and association rule analysis can better find the common prescription drug pairs.The two classification methods of random forest and artificial neural network can well distinguish the six TCM syndrome types of depression syndrome,which have great application significance in the classification and recognition of TCM syndrome types.According to the confusion matrix,the overall accuracy of random forest was 89.44%,and the overall accuracy of artificial neural network was 96.03%,indicating that the classification effect of artificial neural network was slightly higher than that of random forest,but the difference was not too large.The non-linear and fuzzy characteristics of its artificial neural network make its classification effect better than that of random forest classification model,which is more suitable for the related research of TCM syndrome classification,and can provide certain reference value for the future research of TCM diagnosis. |
PDF Full Text Request