Carbohydrate Recommendation for Type-1 Diabetics Patient Using Machine Learning

Abstract

Diabetes is a chronic illness that develops when the blood glucose level is elevated above normal. Diabetes has a variety of reasons, making diagnosis and treatment more difficult than necessary. A patient’s treatment can benefit greatly from a healthy diet. It is important to keep the diet under control so that it doesn’t include an excessive amount of carbohydrates. This study offers assistance in this case by creating a mobile application and website that can suggest a meal item based on the patient’s needs. For this construction, a dataset with basic data about more than fifty different food items is taken from Kaggle. This dataset is then preprocessed utilizingstandardization and encoding methods. To create two Machine Learning (ML)models, two different ML algorithms were applied. In this study, the K Nearest Neighbor (KNN) and Naïve Bayes (NB) algorithms were used. The models are subsequently trained using the preprocessed dataset. The models are also put to the test to see which one forecasts the patient’s ideal food item the most accurately. The NBalgorithm is the best method that may be used for carbohydrate recommendation, according to the testing of these models. This model’s accuracy is 93.12%.The model is therefore installed in the firebase. Another database that contains the patient’s real-time readings is linked to the firebase software as well. The best meal item with the right amount of carbohydrates is then given by the doctor through the website. A food proposal is provided to the patient’s mobile phone together with information like the values of the vital metrics. Based on the patient’s vital signs and required carbohydrate intake, the ML system particularly selects this meal item.