Prediction based on Classifica​tionPartit​ionedModel

Hi all,

I have a predictor matrix X and binary response y (1000 observations) and want to use support vector machine (or other machine learning techniques built in Matlab, i.e., fitctree, fitcdiscr, fitcknn, fitcnet) to train the classifier based on 10-fold cross-validation.

My idea is to use 1-999 observations for cross-validation training and testing, and use the best classifier to predict a single out-of-sample y based on 1000th X. How can I do that?

Without cross-validation, I can simply use predict(.) function in Matlab to predict a single y based on 1000th X. However, this is not allowed when cross-validation is used. For a ClassificationPartitionedModel, the function kfoldPredict(.) should be used. The problem is, I am not allowed to specify the X when using kfoldPredict.

Is there anyone know the answer?

Many thanks.

Answer 1

[Vinayak]

Kindly refer to the following MATLAB Answers post which addresses a similar query

https://www.mathworks.com/matlabcentral/answers/477284-how-to-predict-unknown-data-with-a-regressionpartitionedsvm-model

[XIAODU XIE]

Thank you so much, it really helps.