Anomaly Detection and Failure Prediction
Many practical tasks in engineering, finance, and medicine can be formulated as a problem of anomaly detection. We can solve it by describing the normal state of the object using, e.g., a one-class classification model. In the presentation, we provide a brief overview of models, methods and new approaches to anomaly detection. Particular attention is paid to the practical use of predictive maintenance methods for various engineering systems: monitoring of technical condition of a power unit, forecasting parameters of the roadway on basis of data from road weather stations, etc.