Journals Proceedings

Abstract

K-Nearest Neighbor (kNN) technique is used in different application areas, as it is very simple, highly efficient and effective. Its main advantage is simplicity but the disadvantages can’t be ignored. Hence many researchers proposed different forms of kNN technique under various situations. Broadly kNN techniques are categorized into structure based and non-structure based (structure less) techniques. The objective of this paper is to introspect the key idea, pros and cons and target data or application area behind every kNN technique. Principal Axix Tree (PAT), Orthogonal Structure Tree(OST) , Nearest Feature Line(NFL), Center Line (CL) ,k-d Tree, Ball Tree, Tunable NN etc. are structure based techniques whereas weighted kNN, Model based kNN, Ranked NN (RNN),Condensed NN, Reduced NN, Pseudo/Generalized NN, Clustered kNN (CkNN) , Mutual kNN ( MkNN), constrained RkNN etc. are structure less techniques developed on the basis of kNN. Structure based techniques reduces the computational complexity and structure less methods overcome the memory limitation. Hence structure based kNN techniques can be applied to small volume of data whereas Non-structure based kNN techniques can be applied to large data set.