How do you correct/transform nominal data that is highly skewed while fitting a random forest?

Question

For simplicity:

Socks: Red 1 Blue 2 Yellow 40 Purple 10000

I have One-Hot Encoded these features

Does this work like imbalance problem? Can you transform ca