A marketing analytics team wants to group their active customers into distinct segments based on historical purchasing frequency, average order value, and product categories. The database has no pre-existing segment tags or labels for these customers. Which branch of machine learning should the team use to discover these natural groupings?

Unsupervised learning is a category of machine learning where models are trained on unlabeled data. The primary objective is to identify underlying patterns, structures, or groupings within the dataset without human intervention or predefined categories. In this customer segmentation scenario, the model analyzes input features—such as purchase history, spending habits, and frequency—and groups similar data points together using clustering techniques like K-Means clustering, DBSCAN, or hierarchical clustering. Because there is no target label (or "ground truth") to guide the training, the system relies entirely on mathematical distances or statistical distributions to determine similarity.

Let's look at why the other options do not fit: - Supervised learning requires labeled training data, meaning each customer record would already need a known segment label. The model would learn the relationship between features and these labels to predict categories for new customers, which is not the case here since no labels exist. - Reinforcement learning involves an agent learning to make decisions by performing actions in an environment to maximize a cumulative reward. It operates on a feedback loop of trial and error (actions, states, and rewards) rather than analyzing static historical datasets for grouping. - Semi-supervised learning uses a small amount of labeled data combined with a large amount of unlabeled data. While it could theoretically be applied if some customer profiles were pre-labeled, the scenario explicitly states there are no predefined labels at all, making unsupervised learning the correct and most direct choice.