I am referring to this example:
Can this tutorial be amended so that the predicted value is of de-normalized value (not normed). Prediction of T deg (C) of value ranging from -1 to 1.5 just doesn’t give the complete picture.
9 out of 10 examples on time series often show the normalization/standardization to facilitate ML training. But very seldom the prediction made of of the original scale.
just save the mean and std of the train_dataset for the temperatures. you should get them with train_df[“T (degC)”].mean() and train_df[“T (degC)”].std().
Then just take the results and do : T_original = T_normed * T_std + T_mean
Thank you for your reply.
Yes, there are many suggestion on the internet to save the value of mean and std of the train_dataset to achieve non-normed prediction.
It would be good, if we have an example how this can be concretely done. How can the code of this example be amended?
I have saved the mean and std for denorm purpose, after obtaining predicted value. The value make more sense now, but how do i verify?