“…Motivated by the establishment of various diagnostic signatures based on REOs to aid clinical HCC diagnosis decision, we designed robust and powerful predictors in this work. The developed predictors hybridized several algorithms, i.e., REOs, mRMR 21 , MRMD 22 , support vector machine (SVM) 23 , 24 , k-nearest neighbor (KNN) 24 , decision tree (DT) 25 , 26 , logistic regression (LR) 26 , extreme gradient boosting (XGBoost) 24 , logistic model trees (LMT) 27 , adaptive boosting M1 (AdaBoostM1) 28 and naïve bayes (NB) 29 . The REOs method was used for feature construction, mRMR and MRMD were used for feature ranking and selection, 2902 secreted genes (genes encoding secreted proteins) collected public database were used for feature filtering, and SVM, KNN, DT, LR, XGBoost, LMT, AdaBoostM1 and NB algorithms were used for classification purposes.…”