Regression for non linear data

Question

I have a non-linear dataset. How can I do regression in orange? Orange does not have polynomial regression widget.

Answer 1

You have to add higher order polynomial terms to your dataset yourself. Accordingly, you should add new columns to your dataset which consists of higher order polynomials with the desired combination of the existing features. It means that if you have already a CSV file which contains two columns for different samples, you have to add your desired extra feature columns which are the polynomial combination of the existing features. Then use simple regression.

Answer 2

Orange has a Polynomial Regression widget in the Educational add-on.

Answer 3

In R: Try regression with a higher-order polynomial by adding I(variable^power) to the regression formula. You don't need to add another column if you use polynomial terms in the lm function. e.g, if you want to add the quadratic of a certain variable called feature:

lm(response_variable~feature + I(feature^2), data=Orange)