The weighted k-nearest neighbors (k-NN) classification algorithm is a relatively simple technique to predict the class of an item based on two or more numeric predictor variables. For example, you ...

For k-NN these are the percentages of closest labels. For the demo data with k=7, the result proba vector is [0.71, 0.29, 0.00]. This means that 0.71 * 7 = 5 of the closest neighbors are class 0, and ...