KNN
Also known as · K-Nearest Neighbours
K-nearest neighbours classifies a point by a majority vote among its closest training points, using Euclidean Distance. There is no training step, but prediction is slow.
When to use
Always scale features first. A very small overfits and a very large underfits; the rule of thumb is , odd to avoid ties. It degrades in many dimensions (Curse of Dimensionality).
See DS Lec 3.