Hi,
I need some help in modeling a linear regression problem. I am trying
to fit a relationship between the dependent variable y and the
independent variables matrix X.
I tried different set of models, and also did some EDA and saw clearly
no linear relationship exist between y and X. I also tried with some
transformations of the variables, robust regressions, ace and avas
(the variance stabilization methods) and a few more. But I don't seem
to get a decent model that can validate a subsample.
After this, I want to try some machine learning algorithms, where I
just input the X and y, the algorithm applies a definite set of
transformations and arithmetic operations (something like genetic
algorithm).
I remember there is a software called GenIQ which does something like
and produces a functional form of X and y. It is a MLE. This software
takes the variables X and Y and the simple arithmetic operations (+,
-, *, / etc) and some transformations (like sin, cos, exp) as input
and evaluates a final expression of Y = f(X).
Is there any such algorithm or a related one in R?
I welcome your comments and any such references to existing algorithms in R.
Thank you,
Nagu
