Base class for time series event detection in the Harbinger framework.
It provides shared state handling and helper methods used by anomaly,
change-point, and motif detectors. Concrete detectors extend this class
and implement their own fit() and/or detect() S3 methods.
Details
Internally, this class stores references to the original series, indices of
non-missing observations, and helper structures to restore detection results
in the original series index space. It also exposes utility hooks for
deviation measures, filter criteria, and candidate selection provided by
harutils().
References
Harbinger documentation: https://cefet-rj-dal.github.io/harbinger
Salles, R., Escobar, L., Baroni, L., Zorrilla, R., Ziviani, A., Kreischer, V., Delicato, F., Pires, P. F., Maia, L., Coutinho, R., Assis, L., Ogasawara, E. Harbinger: Um framework para integração e análise de métodos de detecção de eventos em séries temporais. Anais do Simpósio Brasileiro de Banco de Dados (SBBD). In: Anais do XXXV Simpósio Brasileiro de Bancos de Dados. SBC, 28 Sep. 2020. doi:10.5753/sbbd.2020.13626