Cluster User Music Sessions

Examensarbete för masterexamen

Please use this identifier to cite or link to this item: https://hdl.handle.net/20.500.12380/202958
Download file(s):
File Description SizeFormat 
202958.pdfFulltext19.27 MBAdobe PDFView/Open
Type: Examensarbete för masterexamen
Master Thesis
Title: Cluster User Music Sessions
Authors: Carlsson, Oscar
Abstract: The ability to serve the \right music for every moment\ is of crucial importance for streaming services like Spotify. To make that a reality, it's necessary to understand what a music moment is, which tracks that belongs to each music moment, and to understand which moments that are most popular amongst users. This thesis presents a project that attempts to cluster user session with descriptive data on tracks that has not been available before. The objective is to evaluate the usefulness of the descriptive data for this problem, and identify the most popular moments for the users. An approach where the problem is modeled as in text document clustering is presented and showed successful. The genre and \mood\ information about each track is the information from the music listening used. A comparison between di erent widely used techniques from document clustering is shown and the best model show results that it managed to capture clusters of what a human would interpret as similar sessions. The results demonstrate that the method and the descriptive data is able to capture music moments of the users and it also shows a way of interpreting the clusters. Examples of identi ed popular moments are clusters named Pop, Indie Rock, Dance & House, Christmas, Classical, Children's music and Comedy
Keywords: Data- och informationsvetenskap;Computer and Information Science
Issue Date: 2014
Publisher: Chalmers tekniska högskola / Institutionen för data- och informationsteknik (Chalmers)
Chalmers University of Technology / Department of Computer Science and Engineering (Chalmers)
URI: https://hdl.handle.net/20.500.12380/202958
Collection:Examensarbeten för masterexamen // Master Theses



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.