Spotify tracks

10,000 tracks from Spotify playlists of six genres, with audio features, a release date and Spotify’s popularity score.

Source	Collected from the Spotify Web API by Kaylin Pavlik and published as a TidyTuesday dataset (21 January 2020). The course file is a random sample of 10,000 of its tracks. Spotify’s terms restrict commercial use.
Rows	One row per track and playlist (11,554 rows): 1,101 tracks are in more than one playlist. For one row per track, use distinct(track_id, .keep_all = TRUE), as the loader below does.
Columns	track_id, track_name, track_artist, track_popularity (0 to 100), release_date, release_year, playlist_genre, playlist_subgenre, and the audio features danceability, energy, key, loudness, mode, speechiness, acousticness, instrumentalness, liveness, valence, tempo, duration_ms.
Target	track_popularity, Spotify’s score in January 2020. It rewards recent plays, so release_year carries recency as well as era. A tenth of the tracks score exactly 0.
Notes	The genre is the playlist’s, not the track’s. No missing values in the features; one track has no name. Six genres, no lyrics, no artist history, no play counts.

Loader

library(tidyverse)
sp <- read_csv("https://ml4ds.com/files/spotify_tracks.csv") |>
  distinct(track_id, .keep_all = TRUE) |>          # one row per track
  mutate(duration_min = duration_ms / 60000)