What are the different approaches to tackle anomaly detection?
Before we dive into Amazon Lookout for Equipment, first, we are going to look at a few definitions. In this section, you will read about the different types of anomalies before getting a high-level overview of the different methods you can use to build your own anomaly detection models.
What is an anomaly?
An anomaly in a time series is usually defined as an observation or sequence of observations that do not follow the expected behavior of the series. For example, you could have point anomalies (in other words, single events that are only recorded at a single timestamp):
Additionally, you might have a sequence of data points that can be viewed as a suspicious event with a longer time range:
These sequences of anomalies are more challenging to identify than point anomalies as they...