On NBEATS article, we have the description af a interpretable version of NBEATS, and on figure 2, we can see interesting results on of the interpretable version.
We can split our dataset on the classes that we have (on WESAD case: baseline, stress and amusement) and run the NBEATS-I on each of those subdatasets
With the trainned model, we can see if we find patterns of each classe based on the seasonality and on the trend of the classes.
NOTE: to avoid reworking, check if we already have an implementation of NBEATS befora take this task
On NBEATS article, we have the description af a interpretable version of NBEATS, and on figure 2, we can see interesting results on of the interpretable version.
We can split our dataset on the classes that we have (on WESAD case: baseline, stress and amusement) and run the NBEATS-I on each of those subdatasets
With the trainned model, we can see if we find patterns of each classe based on the seasonality and on the trend of the classes.
NOTE: to avoid reworking, check if we already have an implementation of NBEATS befora take this task