Create LDA 2D and 3D plots

16 次查看（过去 30 天）
Emily Pendleton
Emily Pendleton 2018-10-12
评论： James Richard 2019-12-17
Hello, I'm trying to perform Linear Discriminate Analysis (LDA) on 2 groups with 88 variables describing the groups. I would like to plot my data along with the line used to discriminate groups. The code below only allows me to plot the line, but not the points of the group. I am thinking this is similar to the 'score' output of PCA, but I can't find the analogous variable for this output. Please help!
MdlLinear = fitcdiscr(data,categories)
K = MdlLinear.Coeffs(1,2).Const;
L = MdlLinear.Coeffs(2,1).Linear;
f = @(x1,x2) K + L(1)*x1 + L(2)*x2;
figure %create 2D plot
di2 = ezplot(f)
figure %create 3D plot
d13 = fsurf(f)
  2 个评论
Mor Guetta
Mor Guetta 2019-1-15
What I did was first plotting the data and using the 'hold on' function adding the line.
I'm new to this so there probably is a better way, but this works just fine.
figure()
scatter(x1,y1); % 1st group
hold on
scatter(x2,y2); % 2nd group
hold on
d = ezplot(f); % LDA margin
James Richard
James Richard 2019-12-17
You could try to use gscatter function instead instead, to make it easier.
gscatter(x1,x2,class,'rb','.',10,'on','x1','x2');
% plot x1 and x2 data which grouped by class
% x1 color is red
% x2 color is blue
% markers are dot with size of 10
% legend is on
% x1 label is x2
% x2 label is x2
hold on
fimplicit(f);
"ezplot is not recommended as it behaves differently under different environments"
https://www.mathworks.com/matlabcentral/answers/391994-what-causes-difference-between-ezplot-and-fplot-for-the-same-function-plotted-below
Use fimplicit instead to plot it.
Btw, there is a fault on your code.
K = MdlLinear.Coeffs(1,2).Const;
L = MdlLinear.Coeffs(2,1).Linear;
It should be the same!
K = MdlLinear.Coeffs(1,2).Const;
L = MdlLinear.Coeffs(1,2).Linear;
% or
K = MdlLinear.Coeffs(2,1).Const;
L = MdlLinear.Coeffs(2,1).Linear;
https://www.mathworks.com/help/stats/classificationdiscriminant-class.html
It could result into wrong boundaries!
Try to debug it yourself in the Workspace Browser to see the difference.

请先登录，再进行评论。

请先登录，再回答此问题。

回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Dimensionality Reduction and Feature Extraction 的更多信息

标签

Community Treasure Hunt

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

Start Hunting!

Translated by