We randomly select #Pos number of negative graphs from each original graph set to create balanced graph datasets, which are available here (NCI_balanced.zip). This dataset was used in [7] for genral ...