Spotify All

spotify
Author

Jake Starkey

Published

November 2, 2023

Question 1

For Question 1, run the following R command to read the CSV file, spotify_all.csv as data.frame, spotify_all:

library(tidyverse)
library(skimr)
spotify_all <- read.csv('https://bcdanl.github.io/data/spotify_all.csv')
rmarkdown::paged_table(spotify_all)

The data.frame spotify_all includes information about Spotify users' playlists.

  • The unit of observation in spotify_all is a track in a music playlist. 🎶

Q1a 🎶


Find the ten most popular song. 🎵

  • A value of a song is defined as a combination of a artist_name value and a track_name value.

  • Who are artists for those ten most popular song?

Q1a <- spotify_all %>% 
  count(artist_name, track_name) %>% 
  arrange(-n) %>% 
  head(10)

Q1a
        artist_name                          track_name   n
1             Drake                           One Dance 143
2    Kendrick Lamar                             HUMBLE. 142
3  The Chainsmokers                              Closer 129
4              DRAM         Broccoli (feat. Lil Yachty) 127
5       Post Malone                     Congratulations 119
6             Migos Bad and Boujee (feat. Lil Uzi Vert) 117
7              KYLE             iSpy (feat. Lil Yachty) 115
8      Lil Uzi Vert                       XO TOUR Llif3 113
9             Aminé                            Caroline 107
10           Khalid                            Location 102

Q1b 🎼

  • Find the five most popular artist in terms of the number of occurrences in the data.frame, spotify_all.

  • What is the most popular song for each of the five most popular artist?

Q1b <- spotify_all %>% 
  group_by(artist_name) %>% 
  mutate(n_popular_artist = n()) %>% 
  ungroup() %>% 
  mutate( artist_ranking = dense_rank( desc(n_popular_artist) ) ) %>% 
  filter( artist_ranking <= 5) %>% 
  group_by(artist_name, track_name) %>% 
  mutate(n_popular_track = n()) %>% 
  group_by(artist_name) %>% 
  mutate(track_ranking = dense_rank( desc(n_popular_track) ) ) %>% 
  filter( track_ranking <= 2) %>%   # I just wanted to see the top two tracks for each artist
  select(artist_name, artist_ranking, n_popular_artist, track_name, track_ranking, n_popular_track) %>% 
  distinct() %>% 
  arrange(artist_ranking, track_ranking)

Q1b
# A tibble: 10 × 6
# Groups:   artist_name [5]
   artist_name    artist_ranking n_popular_artist track_name  track_ranking
   <chr>                   <int>            <int> <chr>               <int>
 1 Drake                       1             2715 One Dance               1
 2 Drake                       1             2715 Jumpman                 2
 3 Kanye West                  2             1065 Gold Digger             1
 4 Kanye West                  2             1065 Stronger                2
 5 Kendrick Lamar              3             1035 HUMBLE.                 1
 6 Kendrick Lamar              3             1035 DNA.                    2
 7 Rihanna                     4              915 Needed Me               1
 8 Rihanna                     4              915 Work                    2
 9 The Weeknd                  5              913 Starboy                 1
10 The Weeknd                  5              913 The Hills               2
# ℹ 1 more variable: n_popular_track <int>

Q1c 🎹

Provide both (1) ggplot codes and (2) a couple of sentences to describe the relationship between pos and the ten most popular artists.

Q1c <- spotify_all %>% 
  group_by(artist_name) %>% 
  mutate(n_popular_artist = n()) %>% 
  ungroup() %>% 
  mutate( artist_ranking = dense_rank( desc(n_popular_artist) ) ) %>% 
  filter( artist_ranking <= 10) 
  
# boxplot
ggplot(Q1c,
       aes(x = pos, y = fct_reorder(artist_name, -artist_ranking)) ) +
  geom_boxplot() +
  stat_summary(
    fun = mean,
    color = 'red'
  )

ggplot(Q1c) +
  geom_histogram(aes(x = pos), binwidth = 1) + 
  facet_grid(fct_reorder(artist_name, artist_ranking) ~ .  , switch = "y") +
  theme(strip.text.y.left = element_text(angle = 0))

  • All are skewed right.

    • Users tend to locate these popular artists' songs early in their playlist.

  • The distribution of pos does not seem to vary a lot across the ten most popular artists.

  • Anything noticeable can be mentioned.

Q1d 🎵

Create the data.frame with pid-artist level of observations with the following four variables:

  • pid: playlist id

  • playlist_name: name of playlist

  • artist: name of the track's primary artist, which appears only once within a playlist

  • n_artist: number of occurrences of artist within a playlist

Q1d <- spotify_all %>% 
  count(pid, playlist_name, artist_name) %>% 
  rename(n_artist = n) %>% 
  arrange(pid, -n_artist, artist_name)
rmarkdown::paged_table(Q1d)