Research On Blood Glucose Prediction Model Based On Swarm Intelligence Optimization Algorithm And XGBoost

Diabetes is a relatively common disease that can disrupt the blood sugar balance in the body.Unstable blood sugar levels can cause a series of complications,and even life-threatening in severe cases.At present,there is no cure for diabetes,but the patient’s blood glucose can be controlled by controlling the input of insulin.Therefore,accurate prediction and control of blood glucose has become an indispensable part of the treatment of diabetes.The dynamic blood glucose monitor can record the patient’s blood glucose value at any time,and then obtain the patient’s historical blood glucose data,which lays the foundation for the data-driven blood glucose prediction model.Using the patient’s historical blood glucose data to predict future blood glucose changes can effectively reduce the occurrence of high and low blood glucose events and improve the patient’s living standards.This paper does research on the following three aspects of blood glucose prediction:The first is the processing of non-stationary and non-linear signals.Since the blood glucose data is non-stationary,large errors will be generated when directly input into the model prediction,and it is difficult to distinguish the performance of the model.Therefore,this paper first uses Kalman filter to smooth the blood glucose data and then compare them separately.The influence of the components obtained from discrete wavelet decomposition and empirical mode decomposition on the accuracy of the model’s prediction is verified,which verifies that adaptive empirical mode decomposition is more suitable for the processing of blood glucose data.The second is to use the swarm intelligence optimization algorithm to optimize the parameters of the nuclear extreme learning machine.The choice of parameters often affects the accuracy of the model.Manual adjustment will increase the workload.The optimal parameters can be obtained by directly using the optimization algorithm,thereby improving the prediction accuracy of the model.This paper applies the sparrow search algorithm to the field of blood glucose prediction,and at the same time verifies the feasibility of the sparrow search algorithm in the field of blood glucose prediction;the third is to use the XGBoost method to screen blood glucose data features,and different features have different effects on the model prediction results.Using XGBoost to sort the importance of features,remove the less important features,and improve the accuracy of blood glucose prediction while reducing the input data.