How to fix multicollinearity when the dataset is too many dimensions?

Question

I want to use a linear model to predict two paremeters Y1 and Y2. My dimension of the input X is large, but the number of samples，nearly 200, is small. I tried normal linear