Anomaly detection is the process of finding outlier values in a series of data. That process assumes you have data that falls within a certain understood range (based on historical data, for example), and that occasional values outside that range happen fairly infrequently.
Supervised anomaly detection allows historical data to be labeled as “normal” and “abnormal,” in order to develop models to apply those labels to new data. Anomaly detection can be applied to unlabeled data in unsupervised machine learning, using the historical data to analyze the probability distribution of values that can then determine if a new value is unlikely and therefore an anomaly. Anomaly detection can be performed on a single variable or on a combination of variables. Some examples of multivariate anomaly detection methods include cluster-based local outlier factor, histogram-based outlier detection, isolation forest, and k-nearest neighbors [link: https://towardsdatascience.com/anomaly-detection-for-dummies-15f148e559c1].
Anomaly detection identifies exceptional data values that can correspond to an event of interest such as equipment failure in a predictive maintenance application. During normal operation, historical data may show that operating temperatures fall within a certain range, and that deviations above that range usually lead to a failure. Based on that analysis, a high temperature reading above a certain threshold can be viewed as a sign of an impending failure and investigated further, prompting a maintenance inspection and the pre-positioning of a replacement part.
C3 AI delivers anomaly detection as part of a portfolio of prebuilt, SaaS enterprise AI applications for a growing number of use cases such as C3 AI Reliability, C3 AI Inventory Optimization, C3 AI Fraud Detection, C3 AI Anti-Money Laundering, and more.
Anomaly detection algorithms are offered as part of the C3 AI® Suite for enterprises want to build their own AI-based applications.