Dear expert,
I want to know how I can do K- fold cross validation in my data set in MATLAB. I have read from MATLAB help, but I do not understand the species - is total no of observation (it means - I have total 38 data set - I have to put total no 38 or how can I define?) K is number of fold - I want to fold 5 fold so in my case it is 5 )
nother is what is meas (train,:)meas(test,:), and species (train ,:) I have 8 data set in test data set and 5 are in validation set and 25 data set are intraining how can I code this in my following coding please help me to understand this
%++++++++++++for Cross validdation+++++++++++++++++++++++++++++++++++++++++
[trainP,valP,testP] = divideind(p,trainInd,valInd,testInd);
[trainT,valT,testT] = divideind(t,trainInd,valInd,testInd);
indices = crossvalind('Kfold',species,5)
cp = classperf(species);
for i = 1:5
test = (indices == i); train = ~test;
class = classify(meas(test,:),meas(train,:),species(train,:));
classperf(cp,class,test)
end
Inside the documentation 'species' is in fact the label/gt for your data.
Thus in your case, for the dataset t where 't' is your dataset and 'tlabel' are the labels
indices = crossvalind('Kfold', tlabel, 5); % for 5 fold cv
for i = 1:5
test = (indices == i); train = ~test;
class = classify(t(test,:),t(train,:),tlabel(train,:));
classperf(cp,class,test)
end
Pay attention that classify is a discriminant analysis classification, you replace by any classification method of your choice
thank you sir ,
sir I have attached my code here but there is error when I run it
the error massage is " indices excess matrix dimension" how can I correct this sir I aspect your great support sir I am student and I am doing my research using artificial neural network please I aspect your great support.
sincerely yours
I upload techinical paper about the k-fold cross validation method to generate re-sampling samples and eight LDFs including SVM. You may undestand my explanation because it is very easy. If you understand it, you can analyze your data by LINGO free version.
@Asya
I think the accuracy variation is due to the 5-fold cross validation partition of the data, since the model is trained for each fold.
for example using the MATLAB function crossvalind to separate the data will return different partition each time you call the function.
Hello;
In my knowledge species is the labels vector and meas is the data.
I wish that will help you.
http://mccormickml.com/2013/08/01/k-fold-cross-validation-with-matlab-code/
above link will help you
Mubashir Ahmad Is the code given in your mentioned link apply on data set having numeric categories?
I am trying to apply on my data set having category 0 and 1.
Please reply. Thanks.
Maryam Butt if you mean there are two classes in your dataset, then there is no problem, you can apply.
you should follow the step by step Guide in that link.
http://mccormickml.com/2013/08/01/k-fold-cross-validation-with-matlab-code/
Below library will help you to build cross validation in matlab with binary classification using LibSVM
https://ch.mathworks.com/matlabcentral/fileexchange/46828-k-fold-cross-validation?requestedDomain=en
Marryam you can use the code i given to you in the second link. It's easy way to classify binary problems.
Following Matlab functions can do K-fold cross validation
cvpartition()
https://in.mathworks.com/help/stats/cvpartition.html
Examples
https://in.mathworks.com/discovery/cross-validation.html
How to do K-fold
https://in.mathworks.com/matlabcentral/answers/339498-how-to-do-k-fold-cross-validation-in-matlab
crossvalind()
https://in.mathworks.com/help/bioinfo/ref/crossvalind.html
Book: Advanced Numerical Methods with Matlab® 2: Resolution of Nonlinear, Differential and Partial Differential Equations
Mubashir Ahmad How can I use cross validation /Leave one out in following example https://in.mathworks.com/help/deeplearning/ug/train-stacked-autoencoders-for-image-classification.html
Many researchers misunderstand the validation and use the wrong k-fold CV extended from LOO.
I proposed the 100-fold CV (Method1).
It is very simple as follows: Copy the data 100 times and it is the validation data and the pseudo-populaton.
Sort it by descending or ascending order and divide 100 training samples. The training samples must be sampling from the validation sample. This approach satisfies the principle of statistics such as the population and sample.
On the other hand, many researchers use ordinary k-fold CV that ignores statistical knowledge.
I evaluate many gene sets such as Small Matryoshkasof 169 microarrays and evaluate those by ethod1.
I can solve the high-dimensional gene data analysis completely and firstly.
My theory is very simple and gene data are measured only expression values of genes.
Thus, you need not special knowledge of cancer diagnosis.
Although engineering researchers such as Machine Learning, Bioengineering, and other researchers published many papers from 19995, all results are useless for real gene diagnosis.
Download my recent papers.
- Badges
- Science topic