Chapter 9: Training and Deploying a Neural Network to Predict Glucose Levels

In this chapter, we will look at a more critical prediction problem: forecasting glucose levels to provide diabetics with early warnings when their insulin or carbohydrates need to be balanced due to their blood sugar levels.

We will also introduce neural networks. We will start by learning how to use a neural network for time series prediction by using the simplest neural architecture: a classic feedforward neural network (FFNN) without any recurrent layers. Though simple, the results obtained by this neural network for glucose level prediction are already quite accurate. However, the network’s performance could be improved by using a more complex network architecture, such as long short-term memory (LSTM), which we will introduce in the next chapter.

Thus, we will use this glucose level prediction case study to briefly explain how neural networks work and how they can be trained. Finally, we...