RCMDE-GMD: Predicting gene ontology terms using differential evolution

In this paper, we present a method called RCMDE-GMD (Rule Construction Method Using Differential Evolution Global Multi-Directed Acyclic Graph). The main objective of RCMDE-GMD is to build multi-label global hierarchical classifiers with the hierarchy of classes represented by a DAG. Here, we compare RCMDE-GMD with HLCS-Multi (Hierarchical Learning Classifier System), a method based on Learning Classifier Systems, and with Clus-HMC (Hierarchical Multi-Label Classification), a method based on the induction of decision trees. In the experiments, RCMDE-GMD outperformed HLCS-Multi using one measure from all the datasets and outperformed Clus-HMC in some conditions.