How do I reduce RMSE in a Random Forest Regressor?

Question

I preprocessed the data, normalized the numerical features, and did one hot encoding for the categorical ones. I end up with a model with R^2=0.7 and RMSE which is 15% of the range of values.
I'm okay with the accuracy but I was wondering if there's a way to reduce RMSE to maybe ~7%?


Let me know please.


Thanks!

Answer 1

The path through getting better results from this point is not smooth. There are several recommendations, such as looking at the records that cause the largest errors and finding out the roots of those large errors. Are they outliers? Do you need more data or features?

The other guidelines are presented in this article.