Bachelorarbeit: Local Outlier Factor for Feature‐evolving Data Streams

Aus SDQ-Institutsseminar
Version vom 3. Mai 2021, 03:59 Uhr von Elena Schediwie (Diskussion | Beiträge)
(Unterschied) ← Nächstältere Version | Aktuelle Version (Unterschied) | Nächstjüngere Version → (Unterschied)
Vortragende(r) Elena Schediwie
Vortragstyp Bachelorarbeit
Betreuer(in) Florian Kalinke
Termin Fr 7. Mai 2021
Kurzfassung Outlier detection is a core task of data stream analysis. As such, many algorithms targeting this problem exist, but tend to treat the data as so-called row stream, i.e., observations arrive one at a time with a fixed number of features. However, real-world data often has the form of a feature-evolving stream: Consider the task of analyzing network data in a data center - here, nodes may be added and removed at any time, changing the features of the observed stream. While highly relevant, most existing outlier detection algorithms are not applicable in this setting. Further, increasing the number of features, resulting in high-dimensional data, poses a different set of problems, usually summarized as "the curse of dimensionality".

In this thesis, we propose FeLOF, addressing this challenging setting of outlier detection in feature-evolving and high-dimensional data. Our algorithms extends the well-known Local Outlier Factor algorithm to the feature-evolving stream setting. We employ a variation of StreamHash random hashing projections to create a lower-dimensional feature space embedding, thereby mitigating the effects of the curse of dimensionality. To address non-stationary data distributions, we employ a sliding window approach. FeLOF utilizes efficient data structures to speed up search queries and data updates.

Extensive experiments show that our algorithm achieves state-of-the-art outlier detection performance in the static, row stream and feature-evolving stream settings on real-world benchmark data. Additionally, we provide an evaluation of our StreamHash adaptation, demonstrating its ability to cope with sparsely populated high-dimensional data.