The Application In The Prediction Of Blood Glucose On Fusion Of ARIMA And BPNN

With the growing number of patients with diabetes, the risk to human health is increasing, stabilizing blood glucose is the main objective of clinical treatment in patients with diabetes, if we can predict blood glucose concentration in patients, The doctor and the patient can take measures to stabilize blood sugar before the hyperglycemia or hypoglycemia occur, that will greatly reduce the harm to patients. The establishment of a relatively high degree of precision of blood glucose prediction model that can provide guidance for doctors and patients has a good application value.At present, there are two aspects on the research of human blood glucose prediction technology: one is only considering the historical value of the glucose without considering the impact of external factors(diet, exercise, medication injection) which affect the blood glucose fluctuation, this method is simple and efficient, but not accurate enough.the other direction is not only considering the historical value of the glucose, but also considering The effects of external factors on blood glucose,at the same time, a large number of Pathology, physiology knowledge is considered, the method is accurate but complex and there is a certain delay.In this essay, we analyzed the key factors that affect the blood glucose and the problems we must solve, and then we research the feasibility that using the integration of ARIMA and BPNN to predict future glucose. Firstly, we analyze the historical value of the glucose, find out the linear law of blood glucose changes, then we use ARIMA to predict future glucose. Secondly we use BPNN to capture the effects of external factors by learning and fitting the input values and error. At last, we can get accurate results by combining the ARIMA values and the BPNN values. At the same time, considering the influence of diet and drug injection, we set the start point of impact as the singular point and put forward an algorithm that can find and process the singular point, when the glucose is affected by external interference, the algorithm can adjust the predicted values in the next period of time to ensure the accuracy of the prediction model.The numerical example based on the data from Henan Provincial People’s Hospital, shows that, the proposed hybrid ARIMA-BPNN model and the algorithm which can handle the influenced by external factors are good to predicted blood glucose and the proposed model outperforms the ARIMA model. The result can provide clinical guidance for doctors and patients. The proposed algorithm handle the influenced by external factors can adjust the predicted values in the next period of time to ensure the accuracy of the prediction model.