Algorithm A1 Pseudo-code for aggregate modeling procedure.

$Z\leftarrow load\_data\left(\right)$ $B\leftarrow 100$ $V\leftarrow 0.10$ $F\leftarrow 3.0$ $seg\leftarrow select\_one\left(Both,Pass,Run\right)$ $ntree\leftarrow 1500$ $DT\leftarrow 0.97$ $Perf[:,B]\leftarrow 0$   for $b\leftarrow 1,B$ do     $[Z\_oos,Z\_mdl]\leftarrow partition\left(Z\right)$     $o\leftarrow sample(Z,V\ast nrow(Z),seg)$     $Z\_oos\leftarrow Z[+o,:]$     $Z\_mdl\leftarrow Z[-o,:]$     $s<-strat\_sample(Z\_mdl,F,seg)$     $Z\_mdl\leftarrow Z\_mdl[s,:]$     $mdl\leftarrow gbm(Z\_mdl(ntree,\dots \left)\right)$     $\widehat{y}\leftarrow predict(mdl,Z\_oos)$     $[FDR,TPR]\leftarrow roc(\widehat{y},y,DT)$     $Perf[:,b]\leftarrow accum(FDR,TPR,\dots )$ end for