Adding Pairs to Subsets that are Found in Data File but Have No Pairs

These are subsets that exist in the data table but have no training and no validation pairs. The model cannot reliably predict matches for subsets for which it has not been trained. You likely require that pairs with all types of records that exist in your data table be reliably predicted by the trained model. Therefore, it is recommended that pairs for all subsets that exist in the data table are added to the training dataset.