Prediction of Patient’s Individual Blood Glucose Levels for Type 1Diabetes Using Panel Data Analysis

Nowadays prediction plays an important role in medicine And helps the betterdiagnosis, treatment and control diseases. Diabetes is a chronic disease that its occurrenceis increasingand was known the most costly endocrine diseases in the world and requires continuous control ofblood glucose to prevent complications such as neurological complications, cardiovascular disease,eye disease, amputations and so on. The aim research is based on the modeling and prediction of bloodglucose in patients with type 1 diabetes as a medical necessity due to the factors that influence them.Materials and Methods: This study using factors affecting the patient's blood glucose, such asfood, physical activity, insulin and so on related to six diabetic type 1patientsin the age rangefrom 12to 38 years who were referred to Kermanshah Diabetes Research Center.the affecting factors wasrecorded by the patients in the fourth part of day for 30 days in questionnaires ,To estimate thecoefficients,were used R software and EVIEWS and panel data methods for each individual And finally was obtained the prediction error for each parts of day.Results: The results showed that the panel data method is a good estimator for the coefficients. Onthis basis,food, insulin and exercise has the highest impact on blood glucose to each individual. Themodel was used to predict the changes blood glucoseAnd the mean absolute error with considering theprevious blood glucose were obtained for each patients 6.451,6.096,8.599,7.368,6.039,5.779,respectively.Conclusion: This study, in addition to identify factors that influence blood glucose, can be a basis forfuture research on predictive modeling with high accuracy,So it can be used to help for treatmentdisease in the future.

Keywords:

time series , modeling and prediction , panel data analysis , type 1 diabetes