umbertogriffo/Predictive-Maintenance-using-LSTM

target==s21 ?

linrio opened this issue · 3 comments

linrio commented

I know that the "cycle" is the time unit. And 21 sensor readings is the features. Now I have a question:

if the ground truth data is not the "cycle", the ground truth data is the sensor of s21(for example). How to utilize LSTM to predict the ground truth data(s21)? If I want to predict the values of s21 from 23 cycle to 30 cycle(sequence time series), how to predict?

such as:
image

Tks!

linrio commented

Now I think that we can utilize the one-hot encoder ?

dmfolgado commented

You only need to change the target variable. You have to ensure that for each X_train window, the y_test which is associate is not the RUL but the next respective value for s21.

  • 👍2
umbertogriffo commented

@dmfolgado it's correct, thank you for having answered for me!