Summary

In this chapter, we have concentrated on two aspects of unsupervised methods for time-series:

• Anomaly detection
• Change point detection

The essence of anomaly detection (also: outlier detection) is to identify sequences that are notably different from the rest of the series. We've investigated different anomaly detection methods, and how several major companies are dealing with it at scale.

When working with time-series, it's important to be aware of changes in the data over time that makes models useless (model staleness). This is called change point detection and drift detection.

We've looked at change point detection in this chapter. In Chapter 8, Online Learning for Time-Series, we'll look at drift detection in more detail.