A real-time deep learning-based system for the blood glucose prediction of the diabetic patients

Diabetes is a prevalent and long-term condition. It is possible to prevent diabetes if it is detected early enough. Insulin risk and severity may be considerably decreased if an accurate early prediction could be made. Diabetes datasets are challenging to predict due to the small amount of data that has been categorised and the potential for outliers (or missing values). The purpose of this study is to predict diabetes in a human or a patient with greater accuracy using various machine learning techniques, such as neural networks. Techniques for machine learning may increase the accuracy of predictions with patient-generated data. We begin by classifying the database and separating it into training and testing sets of information, which we then analyse separately. The CNN-LSTM combination is seen to yield the best results, with an accuracy rate of 96.5%. An accuracy of 91.56% is achieved in the testing data. Improved accuracy of the categorisation is achieved by imposing an LSTM model for diabetes prediction.