What type of data is primarily used in supervised learning?

Supervised learning primarily utilizes labeled data, making it essential for training the model. In this learning paradigm, the algorithm is provided with input-output pairs, where each input data point is associated with a corresponding label or output value. This allows the model to learn the mapping from inputs to outputs and to make predictions on new, unseen data.

The presence of labels is crucial because they serve as references that guide the model in identifying patterns and relationships within the data. For example, in a supervised learning task such as classification, the labels represent categories that help the model recognize to which class new examples belong. Similarly, in regression tasks, the labels are continuous output values that the model aims to predict.

This is in contrast to the other options. Unlabeled data, for instance, lacks this output information and is associated with unsupervised learning techniques. Time-series data may be used in supervised learning but isn't a defining characteristic of it, as time series can also be analyzed using unsupervised methods or require structured label definitions. Raw data refers to unprocessed input data that could be labeled or unlabeled but does not inherently define the learning type. Thus, labeled data is fundamental for the success of supervised learning.