Retrieve the data set of out-of-bag.

Mdl = TreeBagger(NumTrees,X,Y,Name,Value)

I developed a random forest model.

I'd like to make internal validation by out-of-bag datasets before external validation.

How can I retrive the data set of out-of-bag?

Answer 1

[TED MOSBY]

Hi @秀和 髙橋

In MATLAB's ‘TreeBagger’, the out-of-bag (OOB) observations are automatically used for internal validation.

Here is how you can do adjust your code:

mdl = TreeBagger(NumTrees, X, Y, 'OOBPrediction', 'On');

If you want to explicitly retrieve the indices or data of the out-of-bag samples for each tree, you can use the ‘OOBIndices’ property of the ‘TreeBagger’ object like below:

To retrieve the indices:

oobIndices = mdl.OOBIndices; % Please note this property is is not available is R2021b, please upgrade MATLAB to leverage this

To retrieve the data:

treeIndex = 1; % Specify the tree number for which you want the OOB data
oobDataX = X(oobIndices(:, treeIndex), :);
oobDataY = Y(oobIndices(:, treeIndex), :);

Please refer to the input arguments and properties section of this link: www.mathworks.com/help/stats/treebagger.html

Hope this helps!