MATLAB and Simulink resources for Arduino, LEGO, and Raspberry Pi

Learn moreOpportunities for recent engineering grads.

Apply Today**New to MATLAB?**

Asked by FIR
on 7 Apr 2012

i have a dataset of 100x6,i want to classify these and find the accuracy using random forest and mlp ,i have classifeid using svm and knn,but dont know how to do with MLP and random forest ,please do help

*No products are associated with this question.*

Answer by Greg Heath
on 11 Apr 2012

For Neural Net classification, see the documentation for patternnet and the classification demo example.

Hope this helps.

Greg

Answer by Ilya
on 7 Apr 2012

If you have Statistics Toolbox and MATLAB 9a or later, you can use TreeBagger. Please read the documentation and take a look at the examples. Follow up with a specific question if something remains unclear.

For MLP, take a look at the Neural Network Toolbox.

Greg Heath
on 11 Apr 2012

See the pattern recognition and classification demos in the Neural Network Toolbox.

Hope this helps.

Greg

Answer by Ilya
on 10 Apr 2012

You can use out-of-bag error as an unbiased estimate of the generalization error. Train TreeBagger with 'oobpred' set to 'on' and call oobError method.

If you insist on using cross-validation, do 'doc crossval' and follow examples there.

FIR
on 11 Apr 2012

i have a code

load fisheriris

groups=species;

cvFolds = crossvalind('kfold', groups, 10); %# get indices of 10-fold CV %# get indices of 10-fold CV

cp = classperf(groups);

for k=1:10

b = TreeBagger(10,meas,species,'oobpred','on');

cp = classperf(groups,b)

end

but i get error as

Error using TreeBagger/subsref (line 884)

Subscripting into TreeBagger using () is not allowed.

Error in classperf (line 219)

gps = varargin{1}(:);

Error in yass (line 10)

cp = classperf(groups,b)

please help

Ilya
on 11 Apr 2012

First, this thread has become convoluted. If you want to post another question, I suggest that you post it as a new question, not as an answer to your own old question.

Second, I suggested that you look at function crossval, not crossvalind. You can cross-validate using crossvalind too, but crossval offers an API that reduces the amount of code you need to write.

Third, whether you choose to use crossval or crossvalind, please take a look at the examples and follow them closely. In particular, 2nd example for crossval here http://www.mathworks.com/help/toolbox/stats/crossval.html shows what you need to do. You would need to replace the function handle classf in that example with a function which has two lines of code in it: 1) Train a TreeBagger on Xtrain and Ytrain, and 2) Predict labels for Xtest using the trained TreeBagger.

Answer by Richard Willey
on 11 Apr 2012

I did a webinar a couple years titled:

"Computational Statistics: An Introduction to Classification with MATLAB". You can watch the recorded webinar online. The demo code and data sets are available on MATLAB Central.

http://www.mathworks.com/matlabcentral/fileexchange/28770-introduction-to-classification

Answer by Agustin
on 17 Nov 2014

I have written the following code to do cross-validation using TreeBagger (I use the fisheriris dataset):

load fisheriris X = meas; y = species;

%data partition cp = cvpartition(y,'k',10); %10-folds

%prediction function classF = @(XTRAIN,ytrain,XTEST)(predict(TreeBagger(50,XTRAIN,ytrain),XTEST));

&missclassification error missclasfError = crossval('mcr',X,y,'predfun',classF,'partition',cp);

I hope it is useful.

## 0 Comments