R squared for SVMR prediction

Report
Question
views

Please explain why do you think this question should be reported?

Report Cancel

I tried making the model with SVMR on my data set having 75% as training set and rest for external prediction the R squared value for calibration and cross validation were quite good, but how to get the R squared value for the Prection data set,
I also tried copying the results into excel and find out the Coefficient of determination but  Later I came to know that for PLS it gives two R squares one is person and the second one,(please tell me how to calculate this second type R square for SVM regression predicted result )

About the Author

Thread Reply

  1. External Admin

    Please briefly explain why you feel this answer should be reported .

    Report Cancel

    To get the R squared value for the Prediction data set in SVMR you will have to make a copy of the predicted values from the test set (Insert-Duplicate Matrix) and copy the reference values into the adjacent column (Ctrl-Shift-C and then Insert copied cells), then make a Scatter plot with the Statistics box.

    The difference between the two R squares for a cross validated model is that the cross validation segments are mean centered inside the PLS algorithm to reflect that the segments may be based on batches, raw materials etc. In the case with random cross validation this difference is often very small. This is the reason why the cross validated R2 actually may be higher than the calibrated R2 although it sounds to be counterintuitive.

    This is why copying the values in the plot directly to Excel does not give the identical R2.

    This extra mean centering calculation is not included in the SVMR, thus there is no way to calculate this figure of merit in an exact way.

Leave an answer