Skip to content

CNN

Model description

Convolutional Neural Networks (CNNs) have been a cornerstone in image recognition tasks but have also found utility in time-series analysis. The main idea behind a CNN is the use of convolutional layers that automatically and adaptively learn spatial hierarchies of features. In the context of CGM data, CNNs can capture spatial (time-adjacent) dependencies in the glucose measurements, making them well-suited for forecasting tasks.

The best hyperparameters after optimization are the following:

Hyperparameter Value
n_conv_layers 3
filters 160
kernel_size 3
activation relu
n_dense_layers 2
dense_size 352

More results and loss curves

Aside from the scores reported in the overview, we show here the Confusion Matrix and a regression example for our network

The cm The ts The loss