Question about suffleing time series data for classification problems

Course Q&A TensorFlow Developer Professional Sequences, Time Series and Prediction
1

Hello,
First, thank you for this amazing course, it is very helpful.

I would like to ask about shuffling. In Week 2, it is mentioned that "Shuffling us to rearrange the data so as not to accidentally introduce a sequence bias. Multiple runs will show the data in different arrangements because it gets shuffled randomly. "

Correct me if I am wrong, this does not apply for time series classification right? Because the problem I am working on is detecting certain patterns in EEG signals (brain electrical activity). If we apply suffleing for classification problems, i think this will change the original order of the data, thus not allowing the model to be able to learn the patterns of interest.

Thanks and regards,

2

Your understanding is correct. Shuffling time series data is a bad idea. But, that’s not what we’re doing here.
Shuffling is done after creating the x to y mappings. So, it’s safe to shuffle.
If you take a particular entry from the shuffled data, you’ll observe no randomness.

2 Likes
3

well noted. thank you so much for your answer!