The Early Warning Model For Abnormal Blood Glucose Concentration Based On Ensemble Learning

With the development of artificial intelligence(AI),its technology has been applied in many fields.In the medical field,AI can process and analyze massive medical and health data,obtain insights through cognitive analysis,serve the government,the health care institutions,pharmaceutical enterprises and patients.It can also realize individualization and evidence-based wisdom medicine,promote the innovation and realize value.In terms of diseases,the proportion of deaths caused by chronic diseases such as cardiovascular diseases and diabetes is much higher than that caused by other types of diseases.Abnormal blood sugar concentration can lead to a series of diseases.The most representative of them is diabetes,a complex chronic disease,it also can lead to a series of complications,and is difficult to cure.Therefore,the prevention of such diseases is particularly important.The aim of this paper is to establish an early warning model of abnormal blood sugar concentration by using artificial intelligence technology to predict blood sugar concentration by analyzing and mining related physiological indicators.So that it can remind people to pay attention to and assist medical staff to put forward prevention and treatment programs.Based on the above purposes,this paper established a inference model of blood sugar concentration by using machine learning technology in the field of artificial intelligence.Firstly,this paper investigated the related research on blood sugar,and put forward some physiological indicators which may be related to blood sugar concentration,including blood routine indicators,biochemical indicators and so on.According to the above indicators,7642 samples were collected,and the features of the extracted samples were analyzed and screened.In order to expand the feature space and increase the capacity of the model,new features were constructed.After a series of characteristic engineering work,the input of the model was determined.For the choice of algorithm in the model,this paper mainly chooses the ensemble learning regression algorithm,and also chooses the linear regression algorithm for the convenience of comparison.In order to optimize and improve the single algorithm model,two fusion modes,average fusion model and cross-Stacking fusion model,were proposed in this paper.The validity of the fusion model proposed in this paper was verified by experiments.In addition,the validity of the new features proposed in the feature engineering was verified by experiments too.Finally,a inference model of blood sugar concentration of the fusion model was established.In summary,this paper has done a lot of work in feature engineering.According to the original data,new features were constructed as the input of the model.At the same time,in order to optimize the model,two fusion methods of the model were proposed.Finally,a inference model of blood sugar concentration was established.In the proposed model,the way of feature construction and fusion model is innovative,and the established model can basically achieve the purpose of early warning of abnormal blood sugar.However,the workload of this model in tuning parameter is relatively small,and it can continue to optimize the model in tuning parameter in the future.