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
A Scalable Recommender System for Automatic Playlist Continuation
University of Skövde, School of Informatics.
2018 (English)Independent thesis Advanced level (degree of Master (One Year)), 10 credits / 15 HE creditsStudent thesis
Abstract [en]

As major companies like Spotify, Deezer and Tidal look to improve their music streamingproducts, they repeatedly opt for features that engage with users and lead to a morepersonalised user experience. Automatic playlist continuation enables these platforms tosupport their users with a seamless and smooth interface to enjoy music, own their experience,and discover new songs and artists.This report details a recommender system that enables automatic playlist continuation;providing the recommendation of music tracks to users who are creating new playlists or curatingexisting ones. The recommendation framework given in this report is able to provide accurateand pertinent track recommendation, but also addresses issues of scalability, practicalimplementation and decision transparency, so that commercial enterprises can deploy such asystem more easily and develop a winning strategy for their user experience. Furthermore, therecommender system does not require any rich and varied supply of user data, instead requiringonly basic information as input such as the title of the playlist, the tracks currently in the playlist,and the artists associated with those tracks.To accomplish these goals, the system relies on user-based collaborative filtering; a simple, wellestablishedmethod of recommendation, supported by web-scraping and topic modellingalgorithms that creatively use the supplied data to paint a more holistic image of what kind ofplaylist the user would like. This system was developed using data from the Million PlaylistDataset, released by Spotify in 2018 as part of the Recommender Systems Challenge, evaluatedusing R-precision, normalised discounted cumulative gain, and a proprietary evaluation metriccalled Recommended Song Clicks, that reflects the number of times a user would have to refreshthe list of recommendations provided if the current Spotify user interface was used tocommunicate them. Over an 80:20 train-test split, the scores were: 0.343, 0.224, and 15.73.

Place, publisher, year, edition, pages
2018. , p. 30
National Category
Computer and Information Sciences
Identifiers
URN: urn:nbn:se:his:diva-15822OAI: oai:DiVA.org:his-15822DiVA, id: diva2:1223679
Educational program
Data Science - Master’s Programme
Supervisors
Examiners
Available from: 2018-06-26 Created: 2018-06-25 Last updated: 2018-06-26Bibliographically approved

Open Access in DiVA

fulltext(1378 kB)31 downloads
File information
File name FULLTEXT01.pdfFile size 1378 kBChecksum SHA-512
f028ee706169397942b4f917ecfe242d60e3a1f837e9038cb0de6fc93da1a4c5e2a7573bd31d0b3270fc1b64cb9d838fed8ba4c2c093c7858ae99376673a42cf
Type fulltextMimetype application/pdf

Search in DiVA

By author/editor
Bennett, Jack
By organisation
School of Informatics
Computer and Information Sciences

Search outside of DiVA

GoogleGoogle Scholar
Total: 31 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 10 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