0
$\begingroup$

I need to predict 2 values from a database. For now, I just predict one value with svm regression and I add a coeff to every prediction like that :

mypred_SUP <- mypred + 0.5
mypred_INF <- mypred - 0.5

Is there a better way to predict two values ?

$\endgroup$

2 Answers 2

1
$\begingroup$

I think you are trying to predict 2 values from your training data: the mean which is mypred for you, and the standard deviation of your prediction.

You should look at the Gaussian Processes, that constructs a predictor with a combination of Gaussian variables. An advantage of such model is that they also predict an estimation of the prediction error, often referred as confidence interval.

There are many available tools that can do that, opensource and commercial. You can maybe start with GPy: https://nbviewer.jupyter.org/github/SheffieldML/notebook/blob/master/GPy/index.ipynb More tools are listed here: http://www.gaussianprocess.org/

$\endgroup$
0
$\begingroup$

You can use Multi Target Regression, The MULAN tool has state-of-the-art algorithms to solve those kinds of problem MULAN. Multi target regression is the term used when there are multiple dependent variables.

$\endgroup$
0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.