I'm searching for a multiple polynomial regression tool like the web-based one found here where you enter the training examples (
X), the observations (
Y) and the max polynomial degree. Then, the program gives you the most suitable equation for this data.
Is there a desktop app or script that can do that?
My training Data (
X) is about 1000+ rows, each of which has 4 features.
What I have tried to do is to create an Octave script that, for each example of those, took the 4 features and converted them into all of the polynomial combinations (of 4 variables) of the given degree, then did a multivariate regression using normal equations.
The problem is, if I do it this way I have all the terms. How can I remove the non-significant terms ?