In a supervised ML model, what do the labels provide for making predictions?

In a supervised machine learning model, labels are crucial as they provide the known outcomes or target variables associated with the input data. These labels represent historical data that the model uses to learn patterns and relationships. By having access to this historical data, the model can map various input features to the correct output labels during the training phase. This is essential for making accurate predictions on unseen data, as the model uses the relationships it has learned from the labeled historical data to infer the likely outcome for new inputs.

The other options refer to different concepts. Testing data is used to evaluate the model's performance after training, feature variables are the inputs provided to the model for making predictions, and performance metrics quantify how well the model is doing, but none of these options serve the same fundamental purpose in the context of learning from labeled data as historical data does.