Table 2. Ranking algorithm for acquired miRNA–mRNA groups. The ranking method R() assigns a score for each group by performing an internal cross-validation.

Ranking Algorithm -R(Xs, M,f,r) Xs: any subset of the input gene expression data X, the features are gene expression values M { is a list of groups produced G() function} f is a scalar: split into train and test data r: repeated times (iteration) res={} for aggregation the scores for each mi Generate Score for each mi and then rank according to the score, Rank(mi): For each mi in M smi=0; Perform r time (here r=5) steps 1-5: 1. Perform stratified random sampling to split Xs into train Xt and test Xv data sets according to f (here 80:20) 2. Remove all genes (features) from Xt and Xv which are not in the group mi (Creat sub data that contains just the genes belongs to group mi ) 3. Train classifier on Xt using SVM 4. t = Test classifier on Xv –calculate performance 5. smi = smi + t; Score(mi)= smi /r ; Aggregate performance res= { Union of Score(mi) } Output Return {Rank(m1),Rank(m2),…,Rank(mp)} which is the sort of the list based on the score value of each group