In the C3_W2_Lab_2_bioc_and_negbio notebook, I noticed that the uncertain labels obtained from the labeller was converted to False (meaning the absence of the disease). Is this strategy recommended for real-world data and disease prediction tasks? I’m trying to understand the implications of converting the uncertain labels to positive or negative labels.
it depends on what basis/criterion these uncertain labels were assigned by the labeller and converted to False. Uncertain labels can be reason due to incomplete features (signs or radiographic features) not confirming any presence of disease. Basically the labeller was not sure of what disease to label that data or image which surely does not confirm false or absence of disease.
You need to check in the criterion of labelling data. Have they mentioned what is considered uncertain labels?
These labels will only create noise or false analysis of your exploration.