Algorithm 1: Overall procedure of rule induction

Input: $t$: trained tree, $d$: distillation rate Output: $\mathbb{L}\left\{l\right\}$: Ordered Rule set consists of $l$ layers

Initialize rule set $\mathbb{L}=\varnothing$ For each $l$ in $Layer$:    For each $r$ in $RF$:       For each $t$ in $tree$:          Induce an ith rule ${ℒ}_i^t$ from a tree t          Calculate rule contribution C of a rule path p          $C\left({ℒ}_i^t\right)=\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$p$}\right.{{\displaystyle \sum }}_{\left(i,j\right)\in p}{W}^t\left(i,j\right)$         (3)          Append rule ${ℒ}_i^t$ and its $C\left({ℒ}_i^t\right)$ to $\mathbb{L}{\left\{l\right\}}_t^r$       End for       Sort rules in $\mathbb{L}{\left\{l\right\}}_t^r$ according to rule contribution       Eliminate some of rules with low contribution by d%    End for End for Return $\mathbb{L}$