Getting the confidence of the prediction?

[jandono]

Hello James,

when using Relevance Vector Regression, is it possible to somehow get the confidence of the predictions? For example lets say that my regression model outputs some predictions, what I want to know is how sure my regression model is that those predictions are true.

Kind regards, ac1dxtrem

[Blair-Young]

I too am interested in this issue, I noticed you have a predict_proba function but I'm unsure how to use it.

[stbedoya]

Hello guys,

I have the same problem with the prediction of probabilities. Could you guys solve your problem?

Thanks!

[Sandy4321]

me too

[Spandyie]

Correct if I am wrong but , the mean and standard deviation can be extracted using instance_variable.sigma_ and instance_variable.m_ , isnt it ?

[JamesRitchie]

m_ and sigma_ are parameters of the weights, not the prediction.

I think if you call predict on the regression class with eval_MSE=True you'll get the standard deviation returned along with the mean as a tuple. I copied this convention from scikit-learn, but they've since changed it to return_std, which makes more sense. If I can get around to fixing up this package I'll check it and change it.

[Sandy4321]

And what is about returning probability for confidence?

[woctezuma]

For your information, there is another package for RVM, following scikit-learn conventions as well:

https://github.com/Mind-the-Pineapple/sklearn-rvm

You can see that return_std is accepted:

Reference: https://sklearn-rvm.readthedocs.io/en/latest/auto_examples/plot_compare_rvr_svr.html