Time Series Anomaly Explainability
Time series is a stream of continuous, sequential, indexed, and timestamped data points often plotted in temporal line charts to correlate trends, detect seasonality patterns, create forecasting, and identify anomalies. Time series data is ubiquitous. Examples of time series data are daily stock prices, weekly COVID-19 confirmed cases, monthly rainfall measurements, and annual sales revenue.
The advent of connected technology, storage affordability, and increasing business demand for insights enables many systems to generate more data than most organizations can consume. According to Statista, (https://www.statista.com/statistics/871513/worldwide-data-created/), only 2% of 64.2 zettabytes produced globally in 2020 was retained into 2021.
Finding anomalies in time series presents significant business values and applies to many real-world use cases. For example, companies proactively monitor manufacturing equipment metrics for industrial predictive...