The K-Nearest Neighbours (K-NN) algorithm works on the fundamental idea that data points with similar features will exist in close proximity in feature space.
This is called the locality assumption.
When making a prediction, K-NN finds the closest training examples (the nearest neighbours) to a new data point and bases its prediction on their values.
Options (A), (B), and (C) contradict this principle because they suggest that similar objects are far, random, or scattered, which goes against the logic of K-NN.
Therefore, the correct answer is that similar objects are close to each other.