Matlab Question - Analysis of Key Components

Question

I have a set of 100 observations, where each observation has 45 characteristics. And each of these observations has an attached label that I want to predict based on these 45 characteristics. So this is an input matrix with a size of 45 x 100 and a target matrix with a size of 1 x 100.
The fact is that I want to know how many of these 45 characteristics matter in my dataset, basically an analysis of the main components, and I understand that I can do this using the matlab function processpca.

Could you tell me how can I do this? Suppose an input matrix xwith 45 rows and 100 columns, and ya vector with 100 elements.

Answer 1

Assuming you want to build a 1x100 vector model based on a 45x100 matrix, I'm not sure that the PCA will do what you think. PCA can be used to select variables to evaluate a model, but this is a somewhat indirect way to assemble a set of model functions. Anyway, I suggest reading both:

Analysis of the main components

and...

Enabling PCA

... both of which provide code in MATLAB that does not require any toolkits.

Answer 2

Have you tried COEFF = princomp(x)?

COEFF = princomp(x) (PCA) n-by-p X , . X , . COEFF p-by-p, , . .

Answer 3

, , MATLAB, . , - .

MATLAB, , WEKA (www.cs.waikato.ac.nz/ml/weka/) RapidMiner (quick-i. ). PCA + .

Answer 4

.

Statistics Toolbox . ,

• Naive Bayes Bagged
• (aka Random Forests)

:

• sequentialfs ( )
• relifF
• "treebagger" .

Optimization Toolbox .

, - MathWorks "Compuational Statistics: MTALAB". -

http://www.mathworks.com/company/events/webinars/wbnr51468.html?id=51468&p1=772996255&p2=772996273

MATLAB Central

http://www.mathworks.com/matlabcentral/fileexchange/28770

, . PCA

Answer 5

You must find the correlation matrix. in the following example matlab finds a correlation matrix with the function "corr"

http://www.mathworks.com/help/stats/feature-transformation.html#f75476