A Study On Diagnosing Prediction Model Of Hypertension Identified As Overabundant Liver-fire Syndrome Based On XGboost Algorithm

ObjectiveTaking hypertension patients as the main research object,the research collected the objective pulse parameters data and filterd out the appropriate parameters and variables,then used XGboost algorithm to build a predictive evaluation model of the overabundant liver-fire syndrome in hypertension patients.The research also explored the characteristic variables and predictive accuracy of overabundant liver-fire syndrome in full samples.To a certain extent,the research solve the key scientific problems in the diagnosis and evaluation about the hypertension of overabundant liver-fire syndrome.Method1.Collecting the patient-related informationAccording to the purpose and the content of the research,combined with the existing related factors of hypertension,we established the case observation table,and collected the relevant information of the patient.The case observation table includes the following related contents:patient informed consent,general personal information,general physical examination,hypertension-related factors,lifestyle-related factors.2.Collecting the objective parameter information of pulse diagnosisThe "DS01-A" pulse diagnosis instrument was used to collect the objective parameter information of pulse diagnosis about the Bilateral of Guanmai,including frequency domain indicators and time domain indicators.The qualified person is responsible for the use of the instrument after training,and strictly observes the instrument operation specification and procedures and the precautions during the collection of pulse diagnosis parameters to ensure the reliability and stability of the information collection process.3.Construction of the diagnosis and prediction model for the hypertension of overabundant liver-fire syndromeThe collected information was entered into the database.Firstly,the inter-group information of hypertensive patients was statistically described between the overabundant liver-fire syndrome and the non-overabundant liver-fire syndrome,then selected the variables.Then the research used the XGboost algorithm to construct a predictive model about the hypertension of overabundant liver-fire syndrome based on the objective pulse parameters and related variables.At the same time,in order to verify the accuracy of the model and to further observe the relationship between the overabundant liver-fire syndrome and objective parameters of pulse diagnosis and related variables,the research explored the construction of a predictive model in the whole sample.Result1.This study included a total of 520 subjects.Among them,262 cases are hypertensive patients,including 131 cases of overabundant liver-fire syndrome,and 131 cases of non-overabundant liver-fire syndrome.In the whole sample,hypertension of overabundant liver-fire syndrome and non-hypertension of overabundant liver-fire syndrome were combined into overabundant liver-fire syndrome(a total of 268 cases),hypertension of non-overabundant liver-fire syndrome and the normal group were combined into non-overabundant liver-fire syndrome(252 cases).2.From the comparison of objective indicators of pulse diagnosis,the hypertension of non-overabundant liver-fire syndrome are obviously higher than the hypertension of overabundant liver-fire syndrome on the frequency domain index(h3),time domain index(t1,t4,W1,and W2),and other indicators(Ass,A)in the left hand,also the h4,t,Ass,Ad,A in the right hand.And in the whole sample,the group of overabundant liver-fire syndrome are higher than the non-overabundant liver-fire syndrome on the h2(right),h1(left),and h2(left),but significantly lower on the h5/hl(right),tl/t(right),t4(left),W1(left),and h3/hl(left).3.In terms of general information,this study found that mentally engaged workers in hypertensive patients were more likely to have the overabundant liver-fire syndrome;But in the whole sample,we found that physical workers were more likely to have the overabundant liver-fire syndrome.In the general physical examination,the abdominal circumference and WHR were significantly higher than the other group in hypertensive patients.In the whole sample,the chest circumference of the overabundant liver-fire syndrome was significantly higher than the other,the clinical significance remains to be further explored.In terms of bilateral blood pressure of the whole group,the systolic blood pressure and pulse pressure difference in the overabundant liver-fire syndrome were higher than those in the other group.Judging from the symptoms of TCM in the whole sample,the overabundant liver-fire syndrome was more severe than the other group in dizziness,headache,irritable irritability,dry mouth,phlegm yellowness,and red eyes.Regarding treatment-related factors,there was a statistically significant difference(p<0.05)in whether or not the use of Chinese herbs and blood pressure fluctuates between the two groups in the hypertensive patients.The proportion of the overabundant liver-fire syndrome is relatively high.In terms of lifestyle-related factors,the study found that among the hypertensive patients,there was a significant difference(p<0.05)in whether there was a decrease in salt and edible oil between the two groups,the results showed that salt increased,cooking oil reduction are more likely to suffer from the overabundant liver-fire syndrome.But in the whole sample,it was found that conscious exercise(exercise for walking,increase the exercise frequency)can reduce the incidence of overabundant liver-fire syndrome,and improve lifestyle habits(including reduced drinking,salt reduction,reduced oil consumption,reduced flesh,increased coarse grains)can also reduce the incidence.4.Firstly,the research used XGboost algorithm to build a predictive evaluation model of the overabundant liver-fire syndrome in hypertension patients,160 variables were selected to select the final characteristic variables.After the calculation,it was determined that the optimal model contained 10 characteristic variables,which were ranked in descending order of importance:impatient and irritability,t4(left),t5/t4(left),h4(right),diastolic pressure(right),t1/t(left),h4/h1(left),t2(left),h5/h1(left),abdominal circumference.The model has a prediction rate of 76.47%for overabundant liver-fire syndrome,and the AUC value is close to 0.70,which has a higher prediction accuracy.Further analysis found that the objective indicators of pulse diagnosis contributed the most to the diagnosis of hypertension of overabundant liver-fire syndrome,and the pulse diagnosis indicators were mostly composite indicators,because the composite indicators could better reflect the characteristics of the pulse,so they were more accurate than single indicators.From the perspective of different parts of the Guanmai on the diagnosis of hypertension of overabundant liver-fire syndrome,the contribution rate of the left Guanmai is greater,which is consistent with the TCM theory.Left Guanmai is related to the liver.In order to further investigate the characteristics and prediction accuracy of the diagnosis about the overabundant liver-fire syndrome,as well as to expand the scope of the prediction model,this study explored to build a predictive evaluation model of the overabundant liver-fire syndrome in the whole sample,and found that the optimal model contains 15 characteristics.The variables were moderate irritation,systolic blood pressure(left),mild irritation,t4(left),reduced meat consumption,reduced salt,WHR,t5/t4(left),A(Left),reduction in oil consumption,conscious exercise,tl/t(right),Ad(left),t5/t4(right),conscious improvement living habits.The model has a prediction rate of 76.92%and a model AUC value of 0.815 for the whole sample,indicating that it has a good prediction accuracy.The rate of contribution of the pulse diagnosis index to the model is also large,and in the characteristic variables of the two models,there are duplications in the symptoms of "irritability," pulse diagnosis variables 14,t5/t4,and tl/t.These variables are of great significance for the prediction of the overabundant liver-fire syndrome.At the same time,it has also been found that the full-sample predictive model has increased five lifestyle-related variables,indicating that the lifestyle has a greater impact on the overabundant liver-fire syndrome in the full-sample.5.This study uses a self-documented program to visualize the waterfall chart to better demonstrate the role and predictive contribution of different feature variables to different research objects in the XGboost model,facilitating targeted prevention and intervention.Conclusion1.In this study,the diagnosis and prediction model of hypertension overabundant liver-fire syndrome was preliminarily constructed through the XGboost algorithm,the model has a relatively high accuracy,and the importance variables was in line with clinical practice,which is important for further improving the diagnostic accuracy of the hypertension overabundant liver-fire syndrome.To a certain extent,it can solve the key scientific issues in the objective diagnosis of TCM syndromes.2.This study combined with pulse diagnosis objective parameters,general information,physical examination,TCM symptoms and other relevant factors to conduct a comprehensive analysis,and build a diagnostic model of hypertension overabundant liver-fire syndrome.This exploratory research method is worth promoting to other Chinese medicine syndrome diagnosis research.