the question is this. When applying the Method of k-nearest neighbors in binary classification, a situation arises when the number of neighbors is an even number. How to solve this problem? After all, if the algorithm returns an answer in the form of a pair (True, False), that is, the object has two neighbors of different classes, then uncertainty is obtained. That is, any even number of K can lead to uncertainty if the neighbors are 50% of one and 50% of the second class.
Asked
Active
Viewed 1,019 times
0
-
Does this answer your question? [What would be predicted class of KNN for an even value of K and in case of a tie?](https://stackoverflow.com/questions/46602175/what-would-be-predicted-class-of-knn-for-an-even-value-of-k-and-in-case-of-a-tie) – Capybara Nov 12 '21 at 16:57
-
Apparently yes. – Blademoon Nov 13 '21 at 09:37