Mining the Dynamics of Music Preferences from a Social Networking Site

Abstract

In this paper we present an application of our incremental graph clustering algorithm (DENGRAPH) on a data set obtained from the music community site Last.fm. The aim of our study is to determine the music preferences of people and to observe how the taste in music changes over time. Over a period of 130 weeks, we extract for each interval user profiles of 1,800 users that represent their music listening behavior. By building and incrementally clustering a graph of similar users, we obtain groups of people with similar music preferences. We label these clusters with genres according to the user profiles of the cluster members. Due to the incremental nature of DENGRAPH we show how clusters evolve over time. Besides the growth and decrease of clusters we observe how new clusters emerge and old clusters die. Furthermore, we show the merge and split of clusters. The results of our experiments indicate that DEN-GRAPH is particularly useful to efficiently detect groups of similar users and to track them over time.

DenGraph‐HO: a density‐based hierarchical graph clustering algorithm

“…Our experiments using a real‐world social network obtained from Last . fm show that handling 2500 graph updates using DenGraph‐IO is about 400 times faster than the re‐clustering with DenGraph‐O (Schlitter & Falkowski , Falkowski ).…”

“…created a collaborative recommendation system for Last.fm based on both the social annotation and friendships between users (Konstas et al ., ). In 2009, we applied DenGraph‐IO on the dataset to analyse the dynamics of music communities (Schlitter & Falkowski , Falkowski ). Geleijnse et al .…”

Community detection in Social Media

“…Comparing these studies with the seminal work by Palla et al (2007), it appears that some consensus is formed regarding the possible transfor- mations that communities may undergo. Figure 9 illustrates six basic transformations that have been identified in a number of studies (Lin et al 2007;Palla et al 2007;Schlitter and Falkowski 2009). Basically, there are three types of transformations: (a) one-to-one (Fig.…”

“…For instance, Lin et al study the evolution of blog communities forming around real-world events by analyzing the temporal correlations of interactions between bloggers of the same community through observation of their mutual links Lin et al (2007). Schlitter and Falkowski (2009) study the community dynamics of an online music listening application (last.fm) by use of their dynamic community detection framework.…”

Research 2.0: Research-supporting features of World Wide Web

“…By analyzing on a dataset from the socialnetworking site Last.fm [20], the researchers want to determine the music preferences of people and how they change over time. They proceed by extracting data from each of the users and building/clustering a graph of similar users.…”