Now that the basics of supervised and unsupervised learning are clear, the next step is seeing how they work in real situations. Many global teams, from startups to large companies, use Huawei Cloud to build AI systems that solve practical problems. These problems often involve images, text, audio, or large tables of data. At Skyloop, we guide customers through these choices by showing how each learning style fits different goals. This part of the series introduces simple, real-world examples that reflect how these methods are commonly used on Huawei Cloud.
Supervised learning works best when you know what you want the system to predict. A good example is image recognition. In retail, a store may want to detect whether a product label matches the product on the shelf. They can take thousands of images and label each one correctly. ModelArts can then train a model that learns to spot mismatches. Another example is customer behavior prediction, where a bank may use labeled data to predict if a customer is likely to respond to a new service. Because the labels already exist, the model learns directly from past outcomes and becomes better with every training cycle.
Unsupervised learning shines when you want to discover structure in your data without defining the labels in advance. One common case is clustering, which can group customers who behave in similar ways. For example, an e-commerce platform may not know how many types of shoppers it has. Using clustering tools in ModelArts, the system can find natural groups such as budget shoppers, frequent buyers, or visitors who only browse certain categories. This helps the business organize marketing plans more effectively. Another example is anomaly detection, where the system finds unusual patterns in data. This is useful in monitoring energy usage, network traffic, or manufacturing output, especially when the team does not know what “unusual” looks like ahead of time.
On Huawei Cloud, these ideas become practical through tools that handle data, training, and deployment. ModelArts allows companies to run supervised models alongside unsupervised ones and compare their results. DataArts Studio helps clean and understand data before any model is trained. Teams can even mix both learning styles in one project. For example, unsupervised learning can first discover groups in the data, and supervised learning can later predict which group a new sample belongs to. At Skyloop, we use this combined approach often, especially when customers are building their first AI solution and still exploring their datasets.
These examples show that choosing the right learning style depends on the question you want to answer. If you want clear predictions and have labeled data, supervised learning is usually the right path. If you want to explore data structure or find unusual behavior, unsupervised learning is a better match. Huawei Cloud supports both with tools that adapt to each business need. In Part 3 of this series, we will look at how Skyloop helps teams combine these approaches in real projects and how Huawei Cloud platforms turn these ideas into full AI workflows from start to finish.
