Question



Discriminating between neighbors in K-nearest neighbors where distance measures are equal

Dr. Ballings,

Suppose we have the following vector of distance measures corresponding to the distance between a new unlabeled data point and a set of
labeled data points: <0,1,2,3,3,3,3>, and suppose we want to classify our new data point using its five nearest neighbors. If half of the
labeled data points with a distance measure of three are of one class and the other half are of a different class, how do we choose between
these neighbors? In general, how do we choose between k-nearest neighbors who have equal distance measures but that have different labels?

Thanks,
Todd





Answers and follow-up questions





Answer or follow-up question 1

Dear Todd,

See here for the answer:
http://www.ballings.co/qa/question.php?nbr=18

Thanks,
Michel Ballings



Sign in to be able to add an answer or mark this question as resolved.