Too Long; Didn't Read
At its heart, the k Nearest Neighbor technique treats the problems it receives as issues of pattern recognition in order to properly classify data. It does this by ‘learning’ the graphical space in which the different categories reside; implicit is the assumption that each classes will be clustered distinctly. Computers learn similarly to living things, based on as much previous data as is available, predictions are made in order to do work more efficiently.