Smart
Alarming Methods: an overview,highlight on statistical
methods. JeanPaul VALOIS, Christophe BLONDEAU, Simplice DOSSOUGBETE, Laurent
BORDES.
La revue MODULAD, numéro 36, Juillet 2007
Abstract:
Methods of Smart Alarming intend to detect as
soon as possible novelty or anomaly in Data Streams. A review
is proposed to highlight the key points of using them. In case
of univariate data, the more suitable method is not the same
as for stationary variable or nonstationary variable. Multivariate
data set are often dealt with using unsupervised learning based
methods, either with factor analysis (mostly PCA) or clustering
algorithms. Each of these methods must be applied in a specific
situation: prior knowledge of possible anomalies should be
needed or not, learning data set can be large sized or not,
and so on. Some examples are outlined. Discussion underlines
the importance of having a prior knowledge of variable behaviour,
and to consider the global flow chart, including eventually
a data preprocessing.
Keywords: Smart alarming, Novelty detection, Anomaly detection.
Download paper: Smart
Alarming Methods: an overview,highlight on statistical
methods
Download slides: Smart
Alarming Methods: an overview,highlight on statistical
methods
