I have a trained SVR model using the caret package in R. I trained the model for numerical prediction. I used the RBF kernel. I can get the coefficients, support vectors, and parameters from the trained model.
uptakemodel <- train(uptake ~ ., data = training, method = 'svmRadial', trControl = ctrl, tuneLength = 10)
uptakemodel
svp <- uptakemodel$finalModel
xmat <- svp@xmatrix
coeffs <- svp@coef
b <- svp@b
sigma <- svp@kernelf@kpar$sigma
How can I use these parameters and coefficients to predict new values, rather than using the predict() function?
I see an answer from a Python SVR model, and have tried the same method in R, but I am not getting the same answer as when using the predict() function.
