| Objective: People with phlegm-dampness constitution are at risk of metabolic syndrome(metabolic syndrome,MS).Health management of people without MS is a new perspective to prevent MS.In this study,case-control study was used to collect information on MS-related risk factors among adult women with phlegm-dampness constitution,and logistic regression was used to screen the risk factors of MS in this population.A MS risk prediction model suitable for this population was established to provide early warning tools for the occurrence of MS for adult women with phlegm-dampness constitution,and individualized health management for related risk factors was carried out.In order to achieve the goal of preventing or delaying the occurrence of MS in phlegm-dampness constitution population.Method: 1.Using case-control study method,439 adult female volunteers with phlegm-dampness constitution were selected for questionnaires and physical examination,and the relevant data were analyzed by SPSS20.0 statistical software.Univariate and multivariate analysis of MS related factors by binary logistic regression analysis screened out variables with statistical significance(p < 0.05),that is,as a risk factor of metabolic syndrome in adult female population with phlegm-dampness constitution.To establish an early warning model of metabolic syndrome in adult female population with phlegm-dampness constitution.2.Using case-control study method,446 adult female volunteers with phlegm-dampness constitution were selected for questionnaire survey and physical examination.The data were pretreated,Python was used as programming language,JetBrains PyCharm was used as programming software,all the processed data variables were trained,the algorithm was improved,and the Logistic Regression logic of sklearn library was used to screen the risk factors.Regression algorithm and stochastic gradient descent algorithm were used to construct an early warning model of metabolic syndrome in adult women with phlegm-dampness constitution.3.The architecture of intelligent diagnosis assistant system is designed,including model training and user interface.Then the data is preprocessed,and the prediction model is trained by using the logic regression algorithm in machine learning.Python is used as the programming language,pycharm is used as IDE(integrated development environment),and the neural network prediction model is constructed by using the logic regression of the sklean library.The weight and intercept parameters are fitted by gradient rising algorithm.Call the parameters after model training: weight and intercept,and then use QT software to make user interface.Results: 1.Univariate logistic regression analysis showed that there were significant differences in MS among adult women with phlegm-dampness constitution in age,body mass index,hypertension,waist circumference,hip circumference,waist-hip ratio,body fat rate,fasting blood sugar,fasting insulin,insulin resistance,triglyceride,family history of diabetes mellitus,family history of hypertension,predominant meat,pressure of life and work,sleeping habits and phlegm-dampness constitution scale score.Multivariate logistic regression analysis showed that age(OR=1.047),hypertension(OR=13.720),waist-hip ratio(OR=3.122),body fat rate(OR=1.097),family history of hypertension(OR=2.407)and irregular sleeping habits(OR=1.476)were important risk factors for MS in adult women with phlegm-dampness constitution.2.Using Python as programming language,JetBrains PyChram as programming software,using logistic regression and gradient ascent algorithm to construct a predictive model of metabolic syndrome in adult women with phlegm-dampness constitution: using 39 factors as input variables,whether it is metabolic syndrome as an output variable,the accuracy rate was 78.02%.The sklearn library's Logistic Regression algorithm and stochastic gradient descent algorithm were used to reconstruct the model.The input and output scalars were unchanged,and the prediction accuracy was 90.42%.Through data comparison,the non-laboratory detection factors with high influence on the model results were selected as input variables.Whether the metabolic syndrome was used as the output variable,the prediction accuracy was 84.67%,the accuracy decreased by 5.75%,and the accuracy decreased.Analysis of the influencing factors of 5.75% proved that hypertension and triglycerides were the most influential factors of metabolic syndrome.Hypertension,triglycerides,BMI,waist circumference,hip circumference,waist-to-hip ratio,upper arm,body fat percentage,fasting blood glucose,total cholesterol,family diabetes,family hypertension,family hyperlipidemia,familial obesity,family gout,family heart Disease,family cerebrovascular disease,family allergic disease,number of meals,bedtime plus meals,dietary timing,diet,smoking/day,alcohol consumption(ml/day),sleep habits,sleep quality,work-life stress intensity whether or not exercise were used as a model input variable,and whether it was a metabolic syndrome as a model output variable,the prediction accuracy was 89.12%.This model is of great significance for the early warning of metabolic syndrome in adult women with phlegmdampness constitution.Conclusion: 1.Adult women with phlegm-dampness constitution are prone to MS,which is associated with hypertension,waist-hip ratio,body fat rate,dyslipidemia,family history of obesity,number of meals and other risk factors.Prevention and control measures should be taken to adjust the Constitution and reduce the incidence of MS in adult women.2.Machine learning can model the complex relationship among risk factors of phlegmdampness constitution in adult female population.Accumulation and training of data can improve the accuracy of model prediction,realize the early warning of metabolic syndrome in this population,and assist clinical diagnosis and treatment.3.According to the research results of study 1 and Study 2,the variables which are easy to measure and obtain in the risk factors of metabolic syndrome of phlegm dampness constitution of adult women are selected.The model design of intelligent diagnosis assistant system is based on the logic regression model of machine learning used in the previous study. |
PDF Full Text Request