I was attempting to analyze the Wisconsin Breast Cancer Diagnostic dataset. Have a couple of questions / doubts.

Per the attached paper, the performance metrics were worse after dimensionality reduction. Why did that happen? Is that because some information was lost when just 1 feature was selected?

So in such cases is it preferred to work with the original number of features? (30 in this case).

I read that KNN suffers from the curse of dimensionality in the presence of higher dimensions. However, in this case we can see that at K = 5 with 30 features, the accuracy is still 93% (which is not bad). So is it fine to use KNN in case of high number of features like this case?

Link to the research paper I'm referring to in the questions above

Breast Cancer Wisconsin (Diagnostic) Data Set from UCI Learning Repository

  • $\begingroup$ Maybe they just selected the wrong features. Maybe they used the wrong method given the data. $\endgroup$ – user2974951 Oct 23 '19 at 7:54

Your Answer

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

Browse other questions tagged or ask your own question.