Anomaly detection, or finding needles in a haystack, is an important tool in data exploration and unsupervised analytic modeling. Anomaly detection also creates a path to supervised modeling by singling out key examples that an analyst can begin to classify as needles or hay.
Those labeled examples are essential for supervised learning, which is much more powerful than unsupervised learning methods like clustering.
Download this paper to learn about the use of anomaly detection as a tool in data exploration and modeling. The paper distinguishes between outliers and anomalies and provides five powerful methods for detecting outliers, which in turn may help identify anomalies.