Plot 3D hyperplane from fitcsvm results.

Hello,

I am trying to figure out how to plot the resulting decision boundary from fitcsvm using 3 predictors. I was able to reproduce the sample code in 2-dimensions found here: https://www.mathworks.com/help/stats/support-vector-machines-for-binary-classification.html#bsr5b6n

I did also look at this answer: https://stackoverflow.com/questions/16146212/how-to-plot-a-hyper-plane-in-3d-for-the-svm-results/19969412#19969412

But it uses functions no longer supported, and I could not locate the equivalent parameters in the new structure.

However I am struggling to grasp how to extend this to three dimensions. Following the same construct but adding z-dimension did not seem to work with contour3.

Below is the sample code of what I am attempting to run. HOME, Hmax, and gap are the prdeictive metrics generated for each observation from separate code, they are all vectors of the same length.

 X = [HOME, gap, Hmax];

 mdl = fitcsvm([X,labels,'OptimizeHyperparameters','auto',...
    'HyperparameterOptimizationOptions',struct('AcquisitionFunctionName',...
    'expected-improvement-plus'));

 d =0.05; % Step size of the grid
 [x1Grid,x2Grid, x3Grid] = meshgrid(min(X(:,1)):d:max(X(:,1)),...
    min(X(:,2)):d:max(X(:,2)), min(X(:,3)):d:max(X(:,3)));
 xGrid = [x1Grid(:),x2Grid(:),x3Grid(:)];

 [ ~ , scores] = predict(mdl,xGrid);

 zGrid = reshape(scores(:,2), size(x1Grid));

 colormap = zeros(size(dense,1),3);
 colormap(labels==1,1) = 1;
 colormap(labels~=1,3) = 1;

 figure,

 scatter3(HOME, gap, Hmax, 10, colormap)
 xlabel 'HOME', ylabel 'Gap Percentage', zlabel 'Hmax'

 hold on

 plot3(X(mdl.IsSupportVector,1),...
    X(mdl.IsSupportVector,2), X(mdl.IsSupportVector,3), 'ko','MarkerSize',10);
 grid on

Answer 1

[Michael McClelland]

Disregard, I now understand the code and was able to update it to function for what I needed. Below is the updated function for any others looking for the same basic framework.

 function [] = svm_3d_plot(mdl,X,group)
 %Gather support vectors from ClassificationSVM struct
 sv =  mdl.SupportVectors;

 %set step size for finer sampling
 d =0.05;

 %generate grid for predictions at finer sample rate
 [x, y, z] = meshgrid(min(X(:,1)):d:max(X(:,1)),...
    min(X(:,2)):d:max(X(:,2)), min(X(:,3)):d:max(X(:,3)));
 xGrid = [x(:),y(:),z(:)];

 %get scores, f
 [ ~ , f] = predict(mdl,xGrid);

 %reshape to same grid size as the input
 f = reshape(f(:,2), size(x));

 % Assume class labels are 1 and 0 and convert to logical
 t = logical(group);

 %plot data points, color by class label
 figure
 plot3(X(t, 1), X(t, 2), X(t, 3), 'b.');
 hold on
 plot3(X(~t, 1), X(~t, 2), X(~t, 3), 'r.');
 hold on

 % load unscaled support vectors for plotting
 plot3(sv(:, 1), sv(:, 2), sv(:, 3), 'go');

 %plot decision surface
 [faces,verts,~] = isosurface(x, y, z, f, 0, x);
 patch('Vertices', verts, 'Faces', faces, 'FaceColor','k','edgecolor', 
 'none', 'FaceAlpha', 0.2);
 grid on
 box on
 view(3)
 hold off
 end