Automatically determine the health of a system, equipped with sensors, by detecting and graphically representing subsequence anomalies in a time series, without prior knowledge of the system.
Applications
- Internet of Things
- Operations monitoring: aeronautics, automobiles, railways
- Industrial production site monitoring
- Control systems such as SCADA
- Health: monitoring physiological parameters
- Finance: fraud detection
- Computer data center operation health monitoring
Competitive advantages
- No prior knowledge of the domain and anomaly characteristics
- No need of labeled instances (unsupervised method)
- Identification of anomalies of varying lengths
- Identification of single and recurrent anomalies of various types
- High accuracy and fast computing method
Intellectual property
- Patent application filled on May 2020
Keywords
Anomalies - Subsequence anomalies - Outliers Time series - Data series
