curf/music_stream_visual

Personal project to create a custom music streaming visualization.

Spotify Streaming Data Visualization

This was a personal project to create a custom music streaming visualization. It was intended to be an aesthetically pleasing visualization that still held much of its information, thus I opted for minimalism and removing typical plotting elements wherever possible.

The focus was on the entirety of music streaming data from 2020. Given 'Artist', 'Track', and playtime ('endTime' and 'msPlayed') I engineered new features to sum the time played for each artist per day.

Then connecting to Spotify's API, I collected the artist metadata (i.e. associated genres) to sum the total time spent listening to particular genres.

See the individual notebooks for source code and more details:

1. 'spotifyData_EDA.ipynb'(initial parsing of the data and summation of streams)
2. 'spotify_API.ipynb' (connecting to Spotify API using spotiPy to collect artist genres - you'll need an authentication by creating a dev. account https://developer.spotify.com/)
3. 'genre_heatmap.ipynb' (source code for custom formatting of the seaborn heatmap plot)

NOTE: Spotify claims there is more granular genre data per song, but in nearly every example I tried it was an empty array (even for extremely popular songs/artists).

Motivation

• Streamgraph / stackgraph paper
  • ('Stacked Graphs – Geometry & Aesthetics')
• Spotify provides easy API access and semi-easy access to personal data
• Expansion beyond Spotify’s “End of Year” summary:
  • Spotify provides users with some “top” genres, songs, and artists they listened to
• Personalized, emotional connection to listening habits over the year
• Christmas/Birthday gift (asking someone to download their data)

Data

• Users can request their personal data; however, it is somewhat “hidden” within settings – just google “download Spotify user data” to find it.
  • Followed by a 2-step authentication, 2-3 day turnaround time and sent via email
• Data files included:
• User metadata
• Followers/Following
• Saved Library
• Playlists
• Search history
• “inferences” – did not investigate too much but it looks to be tags for a user for advertisements

Final Image

IMAGE KEY: Linear time top to bottom, each column is a genre, each row is one day (months on left side), color bar indicates hours of the genre listened on that day.

Turning Data into Art (i.e. dataisbeautiful)

Bringing the visualization into the real-world in a dignified manner (not just my $50 laser printer) in an aesthetically pleasing way that someone would want to mount it on the wall just like a traditional piece of art.

A lot of formatting to get it just right but was easy to save in the exact size and quality needed for printing. Shout-out to Fireball Printing (Philadelphia, PA) for helping me get everything just right!

Final details:

• 15x5 print on Hahnemuhle Baryta paper (semi-gloss finish)
• 16x6 mat with 0.5” boarder
• Walnut frame