Perform feature extraction and modelling without the use of events and epochs?

šŸ’¬ Support & Discussions
,
1

Hi richard,

  • operating system: Windows 10
    Trying to perform similar thing with different column.
    The current event column that Iā€™m trying to use contains scores after event/class change based on the performance.
    events = mne.find_events(raw,consecutive=True,initial_event=True,shortest_event = 1)

150 events found
Event IDs: [ 9 11 12 14 ā€¦ 83 84 87 88 100]

  1. The event column only registered values for 200 instances (when the events ended) but EEG have instance up to 30000. Therefore the remaining have no values. And also both are in separate file and have different time stamps.
  2. Is there any way to perform feature extraction and logistic regression after removing bad stretches without the use of events and epochs.
  3. How do I round of these events/scores to 10? So we can have 10 variables instead of 40.

Thanks

2

Hi drammock,

Question to understand.
Total number of instance / time stamps of eeg signal recorded is around 500000.
Each of the 500000 have label/classification assigned to it.
When used mne.find_events only 150 are getting detected.
Is this 150 events more than enough to train and test a ML model on it for classification?
With is 150 the model always produces 100% accuracy.
Is there something wrong with approach?

  1. Is there any way to perform feature extraction and logistic regression after removing bad stretches without the use of events and epochs in mne.

Thanks

3

Could you ask your question in a way so it contains all the information one needs to understand what you are trying to do, what you have tried so far (posting your code is the golden standard), and the specific issue are you running into? Specific people will not be available at all times (I see this question has not seen any responses for 18 days). Hence, the spirit of the forum is to try to format your question in such as way to allow anyone to answer your question. You may get quicker answers that way.

2 Likes