|
You know, for traditional regression-based mode,prior assumptions make it far away from reality though simplify the problem and undermine the prediction accuracy. Anyway, it is consistent with causual relationship where economists care about a lot.
RF is supposed to establish a black-box machine. We input everything it requires and it outputs what we need. Compared to statistical or economic theory, efficient algorithm with low computational cost, neat/clean code, and optimized prediction results are much more concerned by machine learning guys
|