Apprendimento supervisionato e non supervisionato

Machine learning splits into two big families: supervised learning, which uses labeled answers, and unsupervised learning, which does not. 🧭

In supervised learning, every row carries a known label, the correct answer the model tries to predict for new data.

Supervised models study pairs of inputs and labels, then generalize that mapping to predict labels they have never seen.

Unsupervised learning has no labels at all. The algorithm must find structure in the data on its own, with nothing telling it what is right.

The most common unsupervised task is clustering: grouping similar rows together so that natural segments emerge from raw data.

Clustering decides who belongs together using a measure of distance, like Euclidean distance between feature values.

Sorting shoppers into segments by what they buy, with no preset categories, is classic clustering in action.

Without labels, there is no simple accuracy to check. You judge clusters by how cohesive and well separated they are.

The same table can feed either approach: add a target column for supervised work, or hold it back to let structure reveal itself.

Reach for supervised learning when you have a clear answer to predict; choose unsupervised when you want to explore unknown patterns.

Both families share one library. Clustering tools sit under sklearn.cluster, ready alongside the classifiers you already know.

Let us pin down what makes a task unsupervised.

Supervised learning predicts known labels; unsupervised learning, like clustering, discovers hidden groups when no labels exist. 🎯