After correcting multicollinearity, there is heteroscedasticity in the final paper, but I don't know how to choose the right to eliminate heteroscedasticity in multivariate linear regression.
There is no need to eliminate it. Heteroscedasticity can be directly calculated by generalized moment method (GMM) or simpler generalized least squares method (GLS).
Eviews should have built-in orders to calculate.