MATLAB Answers

Binning a 3D scatter plot

32 views (last 30 days)
KF on 15 Apr 2021 at 11:57
Commented: Adam Danz on 16 Apr 2021 at 17:34
I have a 3d scatter plot that looks as follows
There are many points in my scatter plot, which makes it impossible to view, so I am looking to find a way to bin the data, and replot it.
What is the best way to do this?
Adam Danz
Adam Danz on 15 Apr 2021 at 21:13
You could compute the binned density of the 3D scatter cloud and show a scatter plot with dot sizes proportional to the density. Alternatively, bubblechart3 may come in handy (r2020b+).

Sign in to comment.

Accepted Answer

Adam Danz
Adam Danz on 15 Apr 2021 at 23:59
Edited: Adam Danz on 16 Apr 2021 at 14:44
3D density plot demos
Here's an assortment of 3D density visualizations though there are many other ways to visualize 3D density.
The script was written on the fly and each demo can be cleaned up and optimized. Consider it a sketch, a scratch pad, a Jackson Pollock painting, or a wall in an italian restaurant after a long day of "see what sticks" tests.
All plots show the same data displayed using different visualizations. Some plots require recent release of matlab (e.g. bubblechart3 - r2020b). Control the number of bins with the nBins variable.
See inline comment for details.
%% Create 3D scatter data
n = 2000;
xyz = zeros(n,3);
xyz(:,1) = randn(n,1).*2+8;
xyz(:,2) = randn(n,1).*5-2;
xyz(:,3) = randn(n,1).*3+4;
ri = randperm(n,round(n*.1));
xyz(ri,:) = xyz(ri,:).*[.8,.4,.5]+[-8,12,6];
%% Compute density
% Put points into 3D bins; xyzBinNum is an nx3 matrix containing
% the bin ID for n values in xyz for the [x,y,z] axes.
nBins = 8; % number of bins
xbins = linspace(min(xyz(:,1)),max(xyz(:,1))*1,nBins+1);
ybins = linspace(min(xyz(:,2)),max(xyz(:,2))*1,nBins+1);
zbins = linspace(min(xyz(:,3)),max(xyz(:,3))*1,nBins+1);
xyzBinNum = [...
discretize(xyz(:,1),xbins), ...
discretize(xyz(:,2),ybins), ...
discretize(xyz(:,3),zbins), ...
% bin3D is a mx3 matrix of m unique 3D bins that appear
% in xyzBinNum, sorted. binNum is a nx1 vector of bin
% numbers identifying the bin for each xyz point. For example,
% b=xyz(j,:) belongs to bins3D(b,:).
[bins3D, ~, binNum] = unique(xyzBinNum, 'rows');
% density is a mx1 vector of integers showing the number of
% xyz points in each of the bins3D. To see the number of points
% in bins3D(k,:), density(k).
density = histcounts(binNum,[1:size(bins3D,1),inf])';
% Compute bin centers
xbinCnt = xbins(2:end)-diff(xbins)/2;
ybinCnt = ybins(2:end)-diff(ybins)/2;
zbinCnt = zbins(2:end)-diff(zbins)/2;
%% Plot raw data
fig = figure();
xyz(:,1), ...
xyz(:,2), ...
xyz(:,3), '.')
title('Raw data')
grid on
box on
%% Plot bubblechart3
xbinCnt(bins3D(:,1)), ...
ybinCnt(bins3D(:,2)), ...
zbinCnt(bins3D(:,3)), ...
density, ...
density, ...
'MarkerFaceAlpha', .3, ...
'MarkerEdgeAlpha', .3)
%% Plot scatter3
xbinCnt(bins3D(:,1)), ...
ybinCnt(bins3D(:,2)), ...
zbinCnt(bins3D(:,3)), ...
density*5+5, ...
density, 'filled', ...
box on
%% Plot marginal histograms
% Histograms are normalized and are not related to
% the axis ticks though their relative height within
% the axes shows the population density (ie, 30% of
% the axis hight means that the bin contains 30% of
% the data.
xyz(:,1), ...
xyz(:,2), ...
xyz(:,3), '.')
grid on
box on
hold on
xcount = histcounts(xyz(:,1),xbins)/size(xyz,1)*range(zlim())+min(zlim());
ycount = histcounts(xyz(:,2),ybins)/size(xyz,1)*range(zlim())+min(zlim());
% x density plotted on xz plane
patch(repelem(xbins(:),2,1), ...
repmat(max(ybins),numel(xbins)*2,1), ...
[min(zbins);repelem(xcount(:),2,1);min(zbins)], ...
% y density plotted on yz plane
patch(repmat(max(xbins),numel(xbins)*2,1), ...
repelem(ybins(:),2,1), ...
[min(zbins);repelem(ycount(:),2,1);min(zbins)], ...
title('Marginal histograms')
%% Plot marginal surface plots
% All 3 surface planes use the same color limit.
xyz(:,1), ...
xyz(:,2), ...
xyz(:,3), ...
10, 'w', 'filled', ...
'MarkerFaceAlpha', .5)
hold on
grid off
xlim([min(xbins), max(xbins)])
ylim([min(ybins), max(ybins)])
zlim([min(zbins), max(zbins)])
% xz plane density
[xm,zm] = ndgrid(xbins, zbins);
xyCount = histcounts2(xyz(:,1),xyz(:,3), xbins,zbins);
% yz plane density
[ym,zm] = ndgrid(ybins, zbins);
yzCount = histcounts2(xyz(:,2),xyz(:,3), ybins,zbins);
% xy plane density
[xm,ym] = ndgrid(xbins, ybins);
xzCount = histcounts2(xyz(:,1),xyz(:,2), xbins,ybins);
box on
maxCount = max([xyCount(:);xyCount(:);xzCount(:)]);
cb = colorbar();
ylabel(cb,'Number of dots')
title('Density Planes')
%% Equate all axis limits
fig.UserData.hlink = linkprop(findobj(fig,'type','axes'),{'xlim','ylim','zlim'});
Adam Danz
Adam Danz on 16 Apr 2021 at 17:34
It was fun!

Sign in to comment.

More Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by