Visualisation of Anomaly Using Mixture Model

Tomoharu Iwata, Kazumi SaitoKES 2004 · OpenReview

Abstract

It is important to conduct on anomaly detection in order to know past big events and to prepare for future crises. We propose a new anomaly detection method that visualizes data in 2- or 3-dimensional space based on the probabilities of belonging to each mixture component and the probability of not belonging to any component. It helps to visually understand not only the magnitude of anomalies but also the relationships among anomalous and normal samples. This may provide new knowledge in data, since we can see data from a different viewpoint. We show the validity of the proposed method by using an artificial time series and an economic time series.
BibTeX
@inproceedings{iwata2004visualisation,
  author = {Tomoharu Iwata and Kazumi Saito},
  title = {{Visualisation of Anomaly Using Mixture Model}},
  booktitle = {KES},
  year = {2004},
  url = {https://openreview.net/forum?id=ggAoLljMhAP},
}

Loading related content...

Figure