Online Univariate Outlier Detection in Final Test: A Robust Rolling Horizon Approach

Abstract

We present an online outlier detection method that is applicable to Final Test. Test limits are constructed based on previous measurements and robust statistics are used to ensure a stable start to the method. We analyze our method using realworld data. Furthermore, we identified some cases which can result in performance degradation, but most experiments show that our method is robust to outliers and able to detect them in an online setting.

Selection of tests for outlier detection

“…In this application we make some modeling choices in order to simplify the problem and do not use all aspects of the more general framework. Furthermore, we consider one specific outlier detection method, which is described in [7]. This method is an online method which updates a robust mean and standard deviation in order to set limits based on deviation from the mean measured in the number of standard deviations.…”

“…This method is an online method which updates a robust mean and standard deviation in order to set limits based on deviation from the mean measured in the number of standard deviations. This method has several parameters, but we only let the model choose the tolerance parameter among three options, for the other parameters we use the default values proposed in [7]. Note that the specific choice of outlier detection method is not the scope of this paper and that any method could have been used.…”