MATLAB: How to classify sample data with two sets of training data using knnclassify.

classificationknnclassify

Hi I am currently new at Matlab and I have been trying to classify data by using knnclassify, so far I understand the concept and the tutorial given by matlab. I wish now to classify a sample of data, say a set of co-ordinates and classify it against two or more training data sets, choosing one of the two training sets (classifying by matrix rather than row). Any help and code would be much appreciated.

Best Answer

I don't have knnclassify(). Perhaps you have an old version, a custom function, or a different toolbox than me. I have knnsearch(). Attached is a demo script showing knnsearch() to classify (x,y) coordinates based on how close they are to training classes. Perhaps it will be useful for you.
% Demo to illustrate K Nearest Neighbor classification for (x,y) coordinate points.
% Setup, clean up, and initialization:
clc;    % Clear the command window.
close all;  % Close all figures (except those of imtool.)
clear;  % Erase all existing variables. Or clearvars if you want.
workspace;  % Make sure the workspace panel is showing.
format long g;
format compact;
fontSize = 20;
markerSize = 10;
lineWidth = 2;
numTrainingPoints = 10;
% Make up coordinates that we'll define as class 1.
trainingCoords1 = rand(numTrainingPoints, 2) + 1;
% Make up coordinates that we'll define as class 2.
trainingCoords2 = rand(numTrainingPoints, 2) * 1.75 + 4;
% Plot both classes of training points.
subplot(2, 2, 1);
plot(trainingCoords1(:, 1), trainingCoords1(:, 2), 'rs', 'LineWidth', lineWidth, 'MarkerSize', markerSize);
hold on;
plot(trainingCoords2(:, 1), trainingCoords2(:, 2), 'b^', 'LineWidth', lineWidth, 'MarkerSize', markerSize);
grid on;
xlim([0, 6]);
ylim([0, 6]);
title('Training data.  Points have known, defined class.', 'FontSize', fontSize, 'Interpreter', 'None');
legend('Class 1', 'Class 2', 'Location', 'northwest');
% Enlarge figure to full screen.
set(gcf, 'Units', 'Normalized', 'Outerposition', [0, 0, 1, 1]);
% Now make up unknown data
numUnknownPoints = 50;
unknownCoords = 3 * (rand(numUnknownPoints, 2) - 0.5) + 3;
% Plot the test data points with as-of-yet unknown class all in black asterisks.
subplot(2, 2, 2);
plot(unknownCoords(:, 1), unknownCoords(:, 2), 'k*', 'LineWidth', lineWidth, 'MarkerSize', markerSize);
grid on;
xlim([0, 6]);
ylim([0, 6]);
legend('As-of-yet Unknown Class', 'Location', 'northwest');
title('Test Data Before Classification', 'FontSize', fontSize, 'Interpreter', 'None');
% Get the classes of the unknown data.
% First collect all the training data into one tall array
trainingCoords = [trainingCoords1; trainingCoords2];
[indexes, distancesOfTheIndexes] = knnsearch(trainingCoords, unknownCoords, ...
  'NSMethod', 'exhaustive',...
  'k', 5,... % Get indexes of the 5 nearest points
  'distance', 'euclidean'); % Regular Pythagorean formula for distance
% Extract the classes
% The way I defined the classes is if the index is <= numTrainingPoints, it's class 1, otherwise it's class 2.
class1Map = indexes <= numTrainingPoints;
class2Map = indexes > numTrainingPoints;
% Now sum the maps horizontally to count the number of each class that was found:
class1Count = sum(class1Map, 2);
class2Count = sum(class2Map, 2);
% Now determine which rows got more class 1 "votes"
% so we'll know whether this coordinate of our unknown test data
% "belongs" to class 1 or class 2.
itIsClass1 = class1Count >= class2Count;
itIsClass2 = class1Count <  class2Count;
% Extract the test coordinates into two arrays, one for class 1 and one for class 2.
class1 = unknownCoords(itIsClass1, :);
class2 = unknownCoords(itIsClass2, :);
% Now plot what we found.  Each class gets the same marker as the training class.
subplot(2, 2, 4);
plot(class1(:, 1), class1(:, 2), 'rs', 'LineWidth', lineWidth, 'MarkerSize', markerSize);
hold on;
plot(class2(:, 1), class2(:, 2), 'b^', 'LineWidth', lineWidth, 'MarkerSize', markerSize);
grid on;
xlim([0, 6]);
ylim([0, 6]);
title('Test Data After Classification', 'FontSize', fontSize, 'Interpreter', 'None');
legend('Estimated to be in Class 1', 'Estimated to be in Class 2', 'Location', 'northwest');
Related Question