Study Azure AI-900 ML Model Types: key concepts, common traps, and exam decision cues.
AI-900 uses machine-learning questions to test whether you understand the shape of the answer, not whether you know algorithm internals. Start with the output type. If the output is a number, category, or unlabeled grouping, the technique usually follows directly.
Supervised learning: Learning from labeled examples where the correct answer is already known during training.
| Technique | Use it when… | Typical business example |
|---|---|---|
| regression | the output is a number | forecast sales, price, or wait time |
| classification | the output is a category or label | fraud or not fraud, approve or deny, churn or stay |
| clustering | the goal is unlabeled grouping | customer segmentation or grouping similar incidents |
| Question | Strongest answer |
|---|---|
| Do labeled target outcomes exist? | supervised learning is possible |
| Do you only want to discover patterns or groups? | unsupervised learning is the stronger first fit |
Classification and regression are supervised because they depend on labels. Clustering is unsupervised because it groups data without predefined labels.
| Term | AI-900 takeaway |
|---|---|
| deep learning | neural-network-based approach used for complex patterns, especially in images, audio, and language |
| transformer architecture | model architecture heavily used in modern language and generative-AI systems |
AI-900 does not expect low-level model math. It expects you to know that deep learning is useful for complex pattern recognition and that transformers are foundational in modern language and generative systems.
| Trap | Better reading |
|---|---|
| choosing classification when the output is a number | use regression for numeric prediction |
| choosing clustering when labels already exist | clustering is for unlabeled grouping |
| assuming all advanced ML is generative AI | deep learning can power many predictive and perception tasks too |