4
$\begingroup$

Background: I am trying to use Orange as to classify if a patient has TB based on their coughing sounds.

In the dataset, there are say 100 patients and for each patient we have 10 coughs. For each cough, we have a full feature vector (170 features).

Giving Orange this dataset and training various learning algorithms is fairly straightforward, but the issue that I have is that Orange will consider each feature vector to be independent of another feature vector, which means it will consider every cough of each patient to be independent, and they aren't.

So my question is: Is there a way to tell orange that all 10 coughs of a patient belong to that patient, and when performing leave-one-out or cross validation methods, all the coughs from each patient should be excluded in each fold?

$\endgroup$
0
$\begingroup$

This is not Orange-specific, but IIUC, you could preprocess your data (e.g. in Python or Excel) to have each of the 10 coughs pertaining to a patient on the same patients line. Thus you would have: 100 rows of patients with each row (10*170 + other patient data) attributes wide.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.