A System for Blood Glucose Monitoring and Smart Insulin Prediction

Diabetes is the disease of the age; the blood glucose levels must be monitored at least three times per day for patients with diabetes mellitus. The method used currently for diabetes monitoring causes discomfort and distress because the patient `s finger needs pricking to get a blood sample. The amount of insulin required in the fumigation process is still determined manually using slow calculations. In this paper, the function of the pancreas in the body is implemented using a smart program. A machine learning (ML) based model is used to track the patient's glucose level and predict the appropriate amount of insulin. We have used artificial neural network (ANN) in our final model since it provided the best prediction of the insulin pattern with mean square error (MSE) of 5.79. We have applied our model on 13 patients of different ages and genders, all with type 1 diabetes. Our final design was developed to be used as a simple and easy to use glucose meter that is compatible with Raspberry Pi to measure the blood glucose level and predict the optimal insulin level considering different conditions. The system basically does glucose level measurement with an average accuracy of 98.7%, patient's data recording, and based on the recorded data the amount of required insulin is predicted with maximum accuracy. We mainly aim to reduce the damage caused by inaccuracy in glucose measurement and determine the appropriate insulin level using a method that is smarter, faster, and more accurate than the manual calculations.