Music Listening Analysis

For years I’ve tracked the music I listen to, either my iTunes playbacks with the scrobbler, or by listening to music on Spotify and synchronizing that with as well. keeps track of all that data about songs, albums, and artists you’ve played, and there’s some nice information that they aggregate and present to you as an user. However, I wanted to do my own kind of aggregations and visualizations, but retrieving my own music listening history data seemed just too difficult as an user. used to have many years ago a Data Export feature, which was abandoned after their redesign. Many modern websites have functionality for users to retrieve their own data archive.

Anyway, I forgot about this for some time, until I found this wonderful website which after only providing it with my username, it gave me a CSV file with my whole listening history in a nice structure.

Then I thought, how will I operate on this data? I didn’t have to think for long, because my favorite tool when doing aggregations and nice stuff with data at work is Apache Spark. Usually we work with Hive tables, but I remembered you can also open other kinds of files.

So I did the reasonable thing, and opened this CSV file with PySpark (these very convenient guides (1, 2) were how I installed the required dependencies for PySpark ready to use with Jupyter/Anaconda).

 |-- uts: integer (nullable = true)
 |-- utc_time: string (nullable = true)
 |-- artist: string (nullable = true)
 |-- artist_mbid: string (nullable = true)
 |-- album: string (nullable = true)
 |-- album_mbid: string (nullable = true)
 |-- track: string (nullable = true)
 |-- track_mbid: string (nullable = true)

That’s the simple schema of the data, and this is just enough information for me to do my queries!

A few interesting (at least for myself) tidbits of data is that according to this, I’ve listened to 14130 unique songs from 2194 artists about 121403 times in the last 8 years!

Some things I want to find out from this data:

  • Listening history for specific songs/albums/artists.
  • Finding songs which I’ve played a lot on a single day or week.
  • Finding albums which I’ve played a lot on a single day or week.
  • Finding artists which I’ve played a lot on a single day or week.
  • Finding top listened songs, albums, and artists.
  • Finding seasonal variations and patterns on the music I listen too.
  • Finding which artists I’m likely to listen together in a brief period of time.

For now I’m still doing some exploratory queries, aggregations, and visualizations on this data, but will update with any interesting things I perform on this data.

I do this, because I want to understand my musical behaviors from the perspective of my own data!