Detecting abrupt changes in missing time series data

Maximilian B. Toller, Bernhard C. Geiger, Roman Kern

Abstract: Highlights•Proposed approach detects abrupt changes in missing data.•Can be combined with different forecasting models (neural or classical).•Constant runtime in every setting, i.e., O(1)<math><mi mathvariant="script" is="true">O</mi><mo stretchy="false" is="true">(</mo><mn is="true">1</mn><mo stretchy="false" is="true">)</mo></math>.•Outperforms several baselines on real-world datasets.