For example, in the case of the data above, there are strange values ??in the SIN curve. This data cannot be judged as an outlier based on the range of values of the SIN curve.
The methods in Model of Outlier can also be used to model outliers.
However, not all outliers can be judged as outliers.
For example, in the case of the data above, there are strange values ??in the SIN curve. This data cannot be judged as an outlier based on the range of values of the SIN curve.
However, in the case of the above example, we can do some manipulation of the data so that the outlier model can be used.
This page is about how to judge the type of outliers that the outlier model cannot use with the original data , by bringing it into the Model of Outlier.
The content of this page is how the outlier model can be used as an anomaly model if devised .
It's on the anomaly model page, but when you actually deal with anomaly phenomena, data alone may not solve the problem. It's an interesting subject for data science , but the road to a solution is rather arduous.
It is relatively easy to define a normal state as compared to an abnormal state that has already happened.
It is difficult to predict anomalies that have never happened before. Even if it is unexpected when it is defined, it would be nice if it could be predicted by extending what was defined as a normal state, but when it is not an extension, there is nothing we can do. We need to anticipate every abnormal situation and prepare a way to detect it.
FMEA and FTA in Reliability Engineering are methods that attempt to make assumptions systematically.
As noted above, the model for abnormal that are not outliers depends on how you define "normal."
I have summarized two types here, but there may be other approaches.
NEXT 
Outliers in time-series data
