umu.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Outlier detection in time series: A comparative study investigating different methods for outlier detection on business performance metrics
Umeå University, Faculty of Science and Technology, Department of Mathematics and Mathematical Statistics.
Umeå University, Faculty of Science and Technology, Department of Mathematics and Mathematical Statistics.
2019 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Abstract [en]

Outlier detection has shown to be a valuable tool in many business areas, as it reveals deviationsin datasets. The purpose of this thesis is to develop an automatic model whichclassies divergent observations as outliers. The model should be able to detect changes involatility and mean, so that employees at Klarna undoubtedly knows when to investigate incertain observations.

Six dierent models of statistical and cluster based techniques have been investigated andanalyzed. The models have been tested on various key performance indexes and validatedby making comparisons to outliers given by Klarna.

The outcome of this project has resulted in an outlier detection model built up as a combinationof the three dierent statistical methods; Bollinger Bands, Average Distance andPruned Exact Linear Time. The model has provided a promising result in alignment withexpected outcomes and is recommended to be used as a monitoring tool at Klarna.

Abstract [sv]

Anomalidetektion har visat sig vara ett vardefullt verktyg inom manga branscher, eftersomdet kan avsloja avvikelser i dataset. Syftet med denna rapport har varit att utveckla enautomatiskt modell som klassicerar divergerande observationer som avvikelser. Modellenbor kunna upptacka volatilitets- och medelvardesforandringar av dataset, sa att anstalldahos Klarna otvivelaktigt vet nar man ska utreda specika observationer.

Sex olika statistiska- och klustring baserade metoder har undersokts och analyserats. Modellenhar testats pa Klarnas interna nyckeltal och validerats genom att jamfora avvikelsersom har diskuterats i samrad med Klarna.

Resultatet av projektet har mynnat ut i en anomalidetektions-modell som byggts upp avtre olika statistiska modeller; Bollinger Bands, Average Distance och Pruned Exact LinearTime. Modellen har pavisat ett lovande resultat och kan rekommenderas att appliceras somett overvakningshjalpmedel.

Place, publisher, year, edition, pages
2019.
Keywords [en]
Outlier detection, time series, statistical methods
National Category
Mathematics
Identifiers
URN: urn:nbn:se:umu:diva-162042OAI: oai:DiVA.org:umu-162042DiVA, id: diva2:1341663
External cooperation
Hilda Thalin
Educational program
Master of Science in Engineering and Management
Supervisors
Examiners
Available from: 2019-08-15 Created: 2019-08-09 Last updated: 2019-08-15Bibliographically approved

Open Access in DiVA

No full text in DiVA

By organisation
Department of Mathematics and Mathematical Statistics
Mathematics

Search outside of DiVA

GoogleGoogle Scholar

urn-nbn

Altmetric score

urn-nbn
Total: 43 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf