We have added anomaly detection as a machine learning type available via the Nexosis API so users can discover anomalies in their data. The process of using anomaly detection is similar to regression and classification, but the prediction domain is now “anomalies” (learn more in the “How It Works” Guide). The algorithms will attempt to find observations that fall outside of what’s normal in a dataset. It can then predict if other observations are anomalous, or outliers, using the generated model.
Sample anomaly detection use cases include fraud detection, network intrusion, and identifying abnormal activity in healthcare data like heart measurements from an EKG.
Check out the Anomaly Detection Quick Start to get started.