Study On The Related Factors Of Postherpetic Neuralgia In Hospitalized Patients With Herpes Zoster In Sichuan Provincial Hospital Of TCM Based On Big Data Analysis

Objective:Using the TreeNet algorithm in the big data and artificial intelligence method,namely the stochastic gradient lifting algorithm,the data of patients with herpes zoster(HZ)in Sichuan TCM Hospital were analyzed and studied,and postherpestic neuralgia(PHN)prediction model was constructed.In order to find the relevant factors of PHN occurrence and the key treatment means to prevent PHN,and provide reference for clinical prevention of PHN.Methods:This study collected data on 1,303 herpes zoster patients hospitalized in the Department of Dermatology,Sichuan Provincial Hospital of Traditional Chinese Medicine from December 2011 to February 2018,listing the basic conditions of each patient(including age,gender,time of onset,location of the disease,Combined disease,etc.,test indicators(including three routine,biochemical tests,infection markers,etc.),treatment(including Western medicine,Chinese medicine,non-drug treatment such as acupuncture,blister extraction,laser therapy,etc.)and treatment end-stage(Whether PHN occurs,PHN is marked as 1 and PHN is marked as 0).All data is then cleaned and the cleaned data is transformed into a number of variables describing the patient information to form an analytical broad table.Finally,using data mining technology,based on whether PHN is classified as the basis,the TreeNet algorithm is used to analyze the data in the wide table,and the PHN prediction model is constructed.Finally,the key factors affecting the occurrence of PHN and the key treatment methods for preventing PHN are obtained.Results:The TreeNet algorithm is used to model the wide-table data,and finally the well-predicted PHN model is obtained.Through the model,the importance ranking of the factors affecting the PHN-related factors and the treatment methods is obtained.And the summary model can be drawn: the accuracy of the model,the receiver operating characteristic curve(ROC)of the test sample reached 0.752,and the ROC of the training sample reached 0.985.By analyzing the confusion matrix under the balance threshold of this model,the overall correct rate of the model can reach the pre-judgment accuracy of 70.27%,which indicates that the model has certain predictability for the occurrence of PHN.Conclusion:A PHN prediction model is constructed by TreeNet algorithm,and the prediction performance of the model is good,which has certain predictability for the occurrence of PHN.Through the importance ranking of model variables,62 variables related to the occurrence and treatment of PHN,including hospitalization days,age,cholinesterase,mean erythrocyte hemoglobin concentration,sodium,uric acid,total carbon dioxide,and Bupleurum,can be obtained.The risk relationship between each variable and PHN provides a reference for clinical prevention of PHN.