Nice work. I would recommend checking for overfit of your model. I've got some background for you if you're interested. I had a model at work that had a ridiculous r squared but was overfit. I did some digging around and found out the original data, unbeknownst to me, had been in triplicate. Creating overfit. Anywho, I'd be happy to take a look at it for you if you're interested. I think Judah Cohen and the Italian guys fell victim to it.
Cheers.