All ML algorithms consume data as an input and are expected to generate insights, predictions, classifications, or analyses as an output. Some algorithms have an additional training step, where the algorithm is trained on some data, tested to make sure that they have learned from the training data, and at a future date given a new data point or set of data for which you desire insights.
All ML algorithms that use training data expect the data to be labeled, or somehow marked with the desired result for that data. For instance, when building a spam filter, you must first teach or train the algorithm on what spam looks like as compared to what normal messages (called ham) look like. You must first train the spam filter on a number of messages, each labeled either spam or ham, so that the algorithm can learn to distinguish between the two. Once the algorithm is...