Feature Selection Package Documentation

k-nearest neighbor (k-NN) is a method of classification that classifies an object based on the attributes and training samples already known. The object is classified by a majority vote of its neighbors, given the attributes most common of its k nearest neighbors.

Method Signature:

function [acc] = knnPrediction( trainX, trainY, testX, testY, k)

Output:
  acc: the accuracy of the classifier at 1 nearest neighbor.

Input:
  trainX: training data, each row is an instance.
  trainY: training data, each column is a class.
  testX: testing data, each row is an instance.
  testY: testing data, each column is a class.
  k: neighbor size.

% Using wine.mat, a sample data set found in the same directory as
% knnPrediction.m.
knnPrediction(X,Y,X,Y,1)

BibTex entry for:

Output-Sensitive Algorithms for Computing Nearest-Neighbour Decision Boundaries

@article{1059748,
author = {Bremner, David and Demaine, Erik and Erickson, Jeff and Iacono, John and Langerman, Stefan and Morin, Pat and Toussaint, Godfried},
title = {Output-Sensitive Algorithms for Computing Nearest-Neighbour Decision Boundaries},
journal = {Discrete Comput. Geom.},
volume = {33},
number = {4},
year = {2005},
issn = {0179-5376},
pages = {593--604},
doi = {http://dx.doi.org/10.1007/s00454-004-1152-0},
publisher = {Springer-Verlag New York, Inc.},
address = {Secaucus, NJ, USA},
}