I used a Random Forest regressor to predict the concrete compressive strength of mixtures with good accuracy. Now, I want to use the Random Forest regressor model to generate possible features when given a specific concrete compressive strength. I will then optimize the costs of the mixtures.
I need to know the parameters in the Random Forest regressor model that is used to make predictions after it has been trained.
Do you mean you want to reverse an RF model e.g. the model structure would be y~xi+...xn, but you would like to predict xi+...+xn~y?
Hi Nanchen, the RF doesn't work as a standard linear model (it's an ensemble of trees), you don't have a linear equation or anything like that. If you want that you should use a multivariate regression. That being said, you can use the RF model to inspect the predicted compressive strength based on the original dataset of predictors (or even a new dataset of predictors!), that would give you an idea of the best mixtures. Of course this only makes sense if the pseudo-r2 of the RF is fairly high.
- Badges
- Science topic