Unsupervised learning is correct because the task is to group unlabeled trip paths by geographic proximity and congestion trends to improve driver dispatching.

Unsupervised learning uses clustering algorithms such as k-means and DBSCAN to group similar GPS traces and traffic features without labels. These clusters can reveal common corridors and congestion patterns that inform scheduling and routing rules, and they can be implemented with platforms such as Amazon SageMaker or with standard machine learning libraries when processing the last 90 days of trip data.

Supervised Learning is incorrect because it requires labeled targets such as predefined route categories and the scenario has no labels to train against.

Reinforcement Learning is incorrect because it focuses on learning policies from rewards through interaction over time and not on discovering structure in a static dataset of past trips.

Semi-Supervised Learning is incorrect because it is intended for problems where a small labeled set is available to guide learning together with many unlabeled examples and this task is purely about clustering unlabeled trip data so semi supervised approaches are unnecessary.

Full AWS Practitioner Certification Question