Y_val in anomoly detection lab

Is the data supposed to be unstructured right? how was the y_val generated? let’s say I have a dataset of temperature for engines, how can I generate y_val like this one? and if it was manually done why do we consider it unsupervised

Anomaly detection uses a small collection of labeled data, in order to set the detection threshold.

Without that, you would only be doing statistical analysis, and you’d have to pick an arbitrary value for what is considered an anomaly (such as three-sigma, etc).