summary: People's social networks significantly predict a song's future popularity, improving machine learning prediction accuracy by 50%. Analyzing data from last.fm, researchers found that friendships and music listening habits are key factors in determining a song's success.
The study highlights the importance of social connections in music trends and suggests that who listens to what influences a song's chances of becoming a hit. The findings could revolutionize hit song prediction models by integrating social network data.
Key Facts:
- Social networks improve hit song prediction accuracy by 50%.
- The study analyzed 2.7 million users, 10 million songs and 300 million plays on last.fm.
- Friendship and user influence are important factors in the dissemination and popularity of a song.
sauce: Complexity Science Hub
Have you ever wondered how your friends influence your musical tastes? In a recent study, researchers from the Complexity Science Hub (CSH) demonstrated that social networks are a strong predictor of a song's future popularity.
By analyzing friendships and listening habits, machine learning predictions improved by 50%.
“Our findings suggest that social factors are as important in the spread of music as artist popularity and genre influences,” says Niklas Reiss from the CSH.
The researchers used common criteria for hit prediction, such as artist recognition and genre popularity, as well as information about listeners' social networks, and improved their hit prediction accuracy from 14% to 21%.
This study Scientific Reportshighlights the power of social connections in music trends.
Digging deeper into the data
The CSH team analyzed data from music platform last.fm, analysing 2.7 million users, 10 million songs and 300 million plays. Reisz said that because users can become friends with each other and share their music tastes, the researchers gained anonymous insights into who is listening to what and who influences who.
For their model, the researchers worked with two networks: one that maps friendships, and another that captures influence dynamics (who listens to a song and who follows it).
“Here too, the nodes in the network are people, but a connection is created when one person listens to a song and, shortly thereafter, another person hears the same song for the first time,” explains Stefan Thurner from CSH.
They looked at the first 200 plays of a new song and predicted the likelihood of the song becoming a hit, defined as a song in the top 1% of most-streamed songs on last.fm.
User influence
The study found that a song's reach depends on the influence of the user's social network. Influential individuals and large interconnected friendship circles fuel a song's popularity. Research shows that information about social networks and the dynamics of social influence can better predict whether a song will be a hit.
“Our results also show that influence flows both ways: people who influence their friends are also influenced by their friends,” explains CSH researcher Vito Servedio.
“In this way, a multi-layered cascade unfolds in a very short space of time, starting with just a few people and quickly reaching many with the song.”
Social power in the music industry
Predicting hits is crucial to the music industry and provides a competitive advantage. While existing models often focus on artist recognition or audience metrics, CSH's research hones in on an often-overlooked social aspect: musical homogeneity, the tendency for friends to listen to similar music.
“What was particularly interesting to us was that this social dimension, this homogeneity of music, has received so little attention until now, even though music has always had a strong social dimension,” Rice says.
Cerner said the study quantifies social influence, offering insights that extend beyond music into areas such as political opinions and attitudes towards climate change.
About this Social Neuroscience and Music Research News
author: Elisa Muth
sauce: Complexity Science Hub
contact: Elisa Muth – Complexity Science Hub
image: Image courtesy of Neuroscience News
Original Research: Open access.
Niklas Reisz et al., “Quantifying the Impact of Homogeneity and Influencer Networks on Predicting Song Popularity.” Scientific Reports
Abstract
Quantifying the impact of homogeneity and influencer networks on predicting song popularity
Predicting the popularity of new songs has become standard practice in the music industry, giving a comparative advantage to those who do it well. To that end, a lot of effort has been put into machine learning predictive models.
In these models, relevant predictive parameters are known to include intrinsic lyric and acoustic features, extrinsic factors (such as the influence and support of publishers), and the artist's past popularity.
Little attention has been paid to the social components of the spread of song popularity. Recently, evidence has been reported of musical homogeneity, that is, the tendency for socially connected people to share musical tastes.
Here, we examine how musical homogeneity can be used to predict song popularity. last An online music platform that allows extracting social links between listeners and their listening patterns.
To quantify the importance of networks in song dissemination and ultimately determine song popularity, we designed a predictive model using musical homogeneity. Impact We explore the parameters and show that incorporating them into state-of-the-art machine learning models improves the prediction of song popularity.
The influence parameter improves the prediction accuracy (TP/(TP + FP)) by about 50%, from 0.14 to 0.21, indicating that social factors in music dissemination play as important a role as the popularity of the artist and the influence of the genre.
