I know that as a question it may seem stupid, but in case it is applying K NN with k = 1 and I have two neighbors at the same distance , what is the best approach to carry out the classification ?


The best approach would really depend on your application and what is important too you.

However, things you can try include:

  • Increase K until there is no tie anymore - if you increase to 2, you will likely have another tie, since they are at the same distance already. So, 3 or higher should do the trick.
  • Include another feature to your classifier - adding another dimension to your space could solve the problem if the values for both these data points are different.
  • Choose another distance metric - you could have a preferred way of measuring distance, but for ties, chose another metric that will break it.
  • Establish a rule for breaking ties, e.g.
    • Pick the class with the most observed data points
    • Randomly assign a class

More information on your application would help to choose a best approach.


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.