In this article, we strengthen the proof methods of some previously weakly
consistent variants of random forests into strongly consistent proof methods,
and improve the data utilization of these variants, in order to obtain better
theoretical properties and experimental performance. In