Random Forests: An Introduction from an Econometrics Perspective
SeriesResearch Master Defense
Date and time
August 23, 2019
10:30 - 11:30
The overlap between Econometrics and Machine Learning is still at its infancy. This paper contributes to connecting the two fields by providing an overview which explains the Random Forests algorithm and its performance combined with highlighted results on the asymptotic properties and the implications of these results for econometric research. The conclusion is that Random Forests performs well in prediction problems, so the algorithm is useful when econometric problems can be fitted into a prediction type framework. However, estimating standard errors, confirming theoretic asymptotic properties and doing inference are not straightforward and still contain a lot of open questions.