Using nearest neighbour as mapper of xy coordinates

This is my first post here so I apologize if this is not right place for this kind of question. I am looking for some tips on using (k)nearest neighbor algorithm as a mapper of hypothetical position of an entity in 2d. space. Training data set contains dimensions (W and H) of space the entities are in, entity type, and entity position (X and Y); for example: W: 10, H: 5, type: someType, X: 7, Y: 3. Different types are constrained to occupy only certain regions of space, eg. someType's X position can be only in range of 0-7, Y in range 0-3, and someType2's X can only be in range of 7-10, Y in range 3-5. My question is how to express an algorithm that based on provided training data (which maps types to correct positions in given space) outputs aproximate position. To be more clear: with input W: 15, H: 7.5, type: someType algorithm should output X: someX, Y: someY. Since same types are clustered I thought that nearest neighbor is right approach, but the problem is not exactly to provide some single label or classification but rather output some compound value X, Y.