Please publish your final equation. I imagine that it will be something like

yt = phi0 + phi1 * MA 3 + phi2 * MA9, where phi1+phi2~=1 and phi0 = 0. This is just roughly guessing the previous price.

This is just not a good application of linear regression at all. It’s actually an auto-regression problem — a specific field which isn’t commonly taught in statistics and ML courses. I suggest you learn more before you publish, because this isn’t a good example of “data-driven”, nor could it be responsibly used by investors.