RapidMiner distinguishes between regular and special attributes. If you already have an example set with prediction and label attribute you have to change the role of these. But typically this roles are set when you import your data and the learning of the model.
You have to do this, because the performance operators expects two special attributes with the role 'label' and 'prediction'. Therefore, insert the "Set Role" operator before the performance operator and select the attribute name (use 'set additional roles' for more than one attribute) and the target role accordingly.