I ran into this kind of problem in my projects and I want to see if there more ways to solve it.

EXAMPLE There is some data about apples and pears and the features are dominant color (red, green) and weight in grams. Some apples and pears in the data are missing the color. I know that I can just ignore and remove those apples and pears but is there another way to solve this problem? Since the data for the color is discrete, one or the other, and not continuous, I can't just plug in the median or average color of the other apples and pears. Let's also say that I can't assume or make educated guesses from the data. How should I solve this problem without removing data?

  • $\begingroup$ Can you do clustering? $\endgroup$ – user2974951 Sep 28 '18 at 6:30
  • $\begingroup$ Yes, but is there another way of filling in the data without altering the rest of it? $\endgroup$ – Ethan Yun Sep 28 '18 at 18:46
  • $\begingroup$ What do you mean without altering the rest of it? $\endgroup$ – user2974951 Sep 28 '18 at 19:45
  • $\begingroup$ When, clustering, I don't want to edit the rest of the features. For example, I don't want to edit the weight or some other feature. If I am not understanding the clustering, please correct me. $\endgroup$ – Ethan Yun Sep 29 '18 at 0:07
  • $\begingroup$ That's not what clustering does, you use clustering to classify data points into clusters, that is get a class for each data point (the data remains unchanged). From there you can check what color is most prevalent in each cluster and use that as a placeholder for missing values. $\endgroup$ – user2974951 Sep 29 '18 at 11:46

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.