Home

Awesome

alt text

Abstract

Music is something universal and has been known for connecting people for a long time.
In the age when diversity in music industry is at its peak, it seems that music has lost its universality and people’s musical taste varies to a great extent.

In this project:

Dataset

For the first part of the project, 'Your top Songs of 2018' Spotify playlists of 25 volunteer users were scraped using Spotify API and Spotipy package in Python. A dataset of 1274 artists and 2207 songs was created consisting of features as artist, genre, track duration, musical characteristics such as acousticness, tempo, liveliness, danceability, key, loudness etc.

For the second part of the project, the track popularity prediction, a dataset from an interactive, audio downloads library called the Free Music Archive (FMA) was used.
In particular, per track metadata such as ID, title, artist, genres and play counts along with audio features provided by Echonest (owned by Spotify since 2014) of 13,129 tracks were used. The data can be accessed at this link.

Our final data story:

The code for data scraping is written in the SpotipyScrapper file, and the main part of the analysis, i.e. preprocessing, network analysis and predictions can be found on the FWBF_Analysis file. The data story we created from this material and slides we used to present our work are just a click away. We hope you'll enjoy exploring it!

Team members:

Acknowledgements

A final word for the amazing working team of the NTDS course at EPFL: our professors- Frossard Pascal, Vandergheynst Pierre, head TA Michael Defferrard and others, thank you! The silhouettes created by James Fenton for the Noun Project were used for the basis of the graphical title of the project.