MktCap as a future

AI Machine Learning
1

I have noticed that MktCap as a future does not work very well, neither in Linear or none-Linear models. If I'm adding MktCap as factor in the ranking system it often boost the results quite a bit.

Below is an example of a simple Linnear Reg model without MktCap as future. First picture is 100% AI factor, second picture is 90% AI Factor and 10% MktCap weight. The last picture is the model with MktCap as future.

Is there other futures/factors that you have noticed that doe not work as good as expected in your "AI" model?

Is there other ways to "force" a future in to a linnear model?

Orginal rank

image
image1410×736 114 KB

Rank 90% Ai, 10% MktCap

image
image1436×771 113 KB

MktCap as a future in the AI model.

image
image1427×858 122 KB

2

Earlier pitmaster talked about how you can use constraints to limit the signs of coefficients of features.