Any support vector regression (SVR) library that gives me the regression model/equation used?

Question

I'm looking for a support vector regression (SVR) library that gives me the regression model/equation that was used to calculate the predicted values. Do you know of any lib to give me a regression model/equation as an output as well as the predicted values?

I used Weka SMOreg, and it works well, but it only gives me back the support vectors and predicted values as an output, but not the actual regression equation that is used to predict the values.

Thanks in advance!

Answer 1

I'm not aware of any such library, and I'd be surprised to find one because support vector machines (1) don't produce neat and simple equations, but (2) aren't really complicated either. If you have the support vectors SVᵢ and their weights αᵢ for an SVM, then the SVM computes the function

∑ᵢ αᵢ K(SVᵢ, x)

where K is the kernel function, eg (〈SVᵢ, x〉 + 1)² for the quadratic kernel where 〈〉 denotes the inner product. See Smola and Schölkopf for details.

Answer 2

I had a similar question, this answer shows how to do it in sklearn by accessing properties within the class (not using official API):

https://stackoverflow.com/a/72420740/9954086

I don't believe it is possible with the official API, however, though not guaranteed, I would not anticipate those variables changing names within the svr class. Would be nice if it was possible with the official API.