Research On Prediction Model Of Gestational Diabetes Mellitus Based On Classifiers Ensemble

In recent years,with the improvement of material living standards and the opening of the national three child policy,as a common type of diabetes,the prevalence of Gestational Diabetes Mellitus(GDM)is also increasing year by year and affecting the health of mothers and fetuses to a certain extent,so it is very important to predict and screen gestational diabetes.The development of machine learning has provided great help in solving problems in various fields.In terms of disease prediction,prior models have been established to identify patients and improve their condition through relevant treatment,making a significant contribution to the diagnosis and treatment of some chronic diseases.Establishing disease prediction models to provide doctors with objective and reliable decision-making support and assist medical personnel in making medical diagnosis,to a certain extent,not only saves medical costs,but also reduces the economic burden of patients.The established disease assisted diagnosis system provides great convenience for patients and medical personnel,and has become one of the research hotspots of artificial intelligence.The data set used in this paper is the medical data set provided by the Tianchi Precision Medical Competition-Artificial Intelligence Assisted Genetic Risk Prediction for diabetes jointly held by Alibaba Cloud and Qing Wutong Gene.The main work of this article is as follows:(1)Data preprocessing: analyzing data information,and filling in missing values with the median of each feature for continuous feature missing values based on the characteristics of the data;For discrete features,different filling methods are used,The missing data of gene variables are filled with a special value of 4,while for the four discrete variables of DM family history,pregnancy,birth and BMI classification,mode is used to fill their missing values;For the processing of outliers,it can be seen from the analysis that there are no outliers in discrete features,while the outliers in continuous features are replaced by means;(2)Feature selection: RFECV algorithm is used to select features after preprocessing,Correlation analysis is performed on 26 selected features,and finally 19 features are selected for inclusion in the model;(3)Establish models:This article builds prediction models using the Logistic Regression(LR)algorithm,Support Vector Machine(SVM)algorithm,Decision Tree(DT)algorithm,Gradient Boosting Decision Tree(GBDT)algorithm,and Random Forest(RF)algorithm respectively.Grid Search CV algorithm is used to adjust model parameters.In the integration model stage,a single classifier is used as a secondary learner to build the model.After comparative analysis,it is found that the model with LR as a secondary learner has better performance.