Anomaly detection concerns the identification of patterns that deviate from normality in a dataset. It is a significant and common problem in a number of fields, and for this reason, a variety of methodologies, both standard and attributable to artificial intelligence, have been developed. The problem is complex due to the extreme variety of dataset and anomalies characteristics. In this paper, a new unsupervised approach to anomaly detection is presented. The algorithm is based on the combination of Principal Component Analysis and Neural Gases which, through the suitable vector quantization of the data, allows the calculation of an effective anomaly score. The method has been tested on numerous benchmarking datasets, reporting excellent results and outperforming other state-of-the-art algorithms for anomaly detection tasks.

Unsupervised Anomaly Detection Combining PCA and Neural Gases

Vannucci M.
;
Colla V.;Zaccara A.;Dettori S.;Laid L.
2024-01-01

Abstract

Anomaly detection concerns the identification of patterns that deviate from normality in a dataset. It is a significant and common problem in a number of fields, and for this reason, a variety of methodologies, both standard and attributable to artificial intelligence, have been developed. The problem is complex due to the extreme variety of dataset and anomalies characteristics. In this paper, a new unsupervised approach to anomaly detection is presented. The algorithm is based on the combination of Principal Component Analysis and Neural Gases which, through the suitable vector quantization of the data, allows the calculation of an effective anomaly score. The method has been tested on numerous benchmarking datasets, reporting excellent results and outperforming other state-of-the-art algorithms for anomaly detection tasks.
2024
9783031624940
9783031624957
File in questo prodotto:
File Dimensione Formato  
Vannucci_et_al_EANN2024.pdf

non disponibili

Tipologia: Documento in Pre-print/Submitted manuscript
Licenza: Copyright dell'editore
Dimensione 18.62 MB
Formato Adobe PDF
18.62 MB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11382/567832
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
social impact