Random Forests: some methodological insights