I wonder if it is possible to retrieve something similar to regression equation from the trained decision tree or random forest for the regression problem? In particular, I want to know how much each attribute impacts on the response factor (not pair-wise, but in total). It is possible to measure feature importance, but can I then use this feature importance values as the equation, e.g. y = 0.2*x1 + 0.5*x2, where 0.2 and 0.5 are normalised feature importance values.
The duplicated answer is provided for random forest, but says nothing about simpler models like decision trees.
