machine learning - Remove Biasing effect on predictive model caused due to unequal ratio of values in one column -

i running predictive model using decision forest regression. there 1 categorical column called "type" has 2 levels "ab" , "cd". data available has more records "ab" compared "cd" (75:25). , therefore model biased more towards "ab" , hence predicts output better "ab". how can reduce biasing effect?

i thinking of converting "type" numeric (0,1) , calculating weighted mean, can please elaborate approach or suggest other approaches, new predictive modeling.

thanks!