Fig. 9: Overview of the Machine Learning Pipeline for Prediction of Blood Glucose Values.

This figure illustrates the machine learning pipeline designed for predicting future blood glucose levels based on various inputs, including continuous blood glucose measurements, food nutrients, heart rate, exercise, and time of day. The pipeline consists of several steps, starting with data preprocessing, followed by the utilization of a recurrent neural network (RNN) comprising LSTM (Long Short-Term Memory) and Dense layers. A The process begins by collecting and preparing the input data, which encompasses continuous blood glucose readings, food nutrient information, heart rate data, exercise data, and time of day. The collected data then undergoes preprocessing, where it is cleaned, normalized, and organized in a suitable format for the subsequent stages. B Next, the preprocessed data is fed into the RNN model, which is composed of LSTM and Dense layers. The LSTM layers are employed to capture temporal dependencies and patterns within the data, enabling the model to understand the sequential nature of blood glucose fluctuations over time. The Dense layers aid in learning complex relationships and extracting relevant features from the input data. C The RNN model is trained to predict the blood glucose level for the next time step. Once the initial prediction is made, it is fed back into the model as an input, allowing the model to generate subsequent predictions for future time steps. This feedback loop enables the model to iteratively refine its predictions and adapt to changing conditions. D The output of the pipeline is a sequence of predicted blood glucose values, which can be used for various applications, such as monitoring and managing blood glucose levels in individuals with diabetes or supporting personalized dietary and exercise recommendations. Overall, this machine learning pipeline offers a systematic approach for blood glucose prediction, leveraging data preprocessing and a recurrent neural network architecture with LSTM and Dense layers to provide accurate and timely forecasts of blood glucose levels.