#
Model of Abnormal but not Outlier

Judge of Outlier
is also used to judge of abnormal value.

But outlier is the part of abnormal values.

For example, in the data of the figure, there is a strange balue.
But this data is not outlier.

But in this case, after some work, we can use the judge of outlier.

##
By Difference of Prediction and Actual

In the case of above, prediction is the value if it is a SIN curve.

If it is the SIN curve, the difference of the abnormal point is large.
And we can use the data as outlier.

###
Variation of Predicion

Prediction is defined by the definition of normal.

In the case of above, diffence between 1 step is small.
We can define the difference to judge of outlier.

##
Difficulty of Definition of Normal

If there is the experience of abnormal,
normal is defined by comparing the abnormal.

FMEA and FTA
are the approaches to study the mechanism totally.

Time Series AnalysisPrediction for Abnormal

Process Analysis for Abnormal Condition

Independent Component Analysis

NEXT Measurement