What you'll learn

Many data scientists approach anomaly detection by reaching straight for an algorithm even before defining what “normal” means, what kind of anomaly they’re after, or whether anomaly detection even fits the problem. In this course, Foundations of Anomaly Detection, you’ll gain the ability to think through an anomaly problem from the ground up, before any modeling begins. First, you’ll explore what anomaly detection is - the core intuition of modeling “normal”, how it differs from classification, regression, and clustering, and the three types of anomalies you’ll encounter. Next, you’ll discover when anomaly detection is the right tool, how it compares to rule-based systems and other approaches, and why defining “normal” is harder than it looks. Finally, you’ll learn how to frame a real-world anomaly problem by combining anomaly type, business risk, and a definition or normal into a clear problem statement, then assessing whether anomaly detection is likely to succeed. When you’re finished with this course, you’ll have the conceptual foundation in anomaly detection needed to recognize where it applies and frame problems that lead to systems worth building.