Spotify Project: Final Analysis and Presentation

4.1. Energy

Energy is a measure from 0.0 to 1.0 and represents a perceptual measure of intensity and activity. Typically, energetic tracks feel fast, loud, and noisy. For example, death metal has high energy, while a Bach prelude scores low on the scale. Perceptual features contributing to this attribute include dynamic range, perceived loudness, timbre, onset rate, and general entropy.

4.2. Valence

A measure from 0.0 to 1.0 describing the musical positiveness conveyed by a track. Tracks with high valence sound more positive (e.g. happy, cheerful, euphoric), while tracks with low valence sound more negative (e.g. sad, depressed, angry).

classify_track_sentiment <- function(valence, energy) {
  if (is.na(valence) | is.na(energy)) {
    return(NA)
  }
  else if (valence >= .5) {
    if (energy >= .5) {
      return('Happy/Joyful')
    } else {
      return('Chill/Peaceful')
    }
  } else {
    if (energy >= .5) {
      return('Turbulent/Angry')
    } else {
      return('Sad/Depressing')
    }
  }
}
track_sentiment = c()
for (i in 1:199){
  
  track_sentiment[i] = classify_track_sentiment(combined_lists[[15]][[i]], combined_lists[[7]][[i]])
  
}
#Adding sentiment column to Combined of four countries
combined_lists<-cbind(combined_lists,track_sentiment)
#Adding Artist Column to Combined Music List
track_audio_combined <- combined_lists %>% 
  select(track.name,track.id,track.artists,track.album.release_date,track.popularity,danceability:tempo,track_sentiment,track.duration_ms)
kable(head(track_audio_combined)) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed", "responsive")) %>% scroll_box(width = "100%", height = "400px")

artist_names = c()
for (i in 1:199){
  
  artist_names[i] <- track_audio_combined[[3]][[i]][[3]]
  
}
combined_lists <-cbind(combined_lists, artist_names)
glimpse(combined_lists)

## Observations: 199
## Variables: 63
## $ playlist_id                        <chr> "37i9dQZEVXbIVYVBNw9D5K", "...
## $ playlist_name                      <chr> "Turkey Top 50", "Turkey To...
## $ playlist_img                       <chr> "https://charts-images.scdn...
## $ playlist_owner_name                <chr> "spotifycharts", "spotifych...
## $ playlist_owner_id                  <chr> "spotifycharts", "spotifych...
## $ danceability                       <dbl> 0.801, 0.743, 0.628, 0.810,...
## $ energy                             <dbl> 0.688, 0.680, 0.725, 0.631,...
## $ key                                <int> 9, 5, 7, 4, 6, 7, 9, 6, 5, ...
## $ loudness                           <dbl> -6.620, -4.344, -7.387, -7....
## $ mode                               <int> 1, 0, 1, 0, 0, 0, 0, 1, 0, ...
## $ speechiness                        <dbl> 0.1130, 0.1030, 0.1100, 0.3...
## $ acousticness                       <dbl> 0.2780, 0.1130, 0.0266, 0.1...
## $ instrumentalness                   <dbl> 8.88e-03, 1.23e-01, 0.00e+0...
## $ liveness                           <dbl> 0.1500, 0.1830, 0.0549, 0.1...
## $ valence                            <dbl> 0.410, 0.694, 0.458, 0.407,...
## $ tempo                              <dbl> 158.003, 180.059, 173.952, ...
## $ track.id                           <chr> "2KlbLTnQ5Wch2oOelW0Y2k", "...
## $ analysis_url                       <chr> "https://api.spotify.com/v1...
## $ time_signature                     <int> 4, 4, 4, 4, 4, 4, 4, 4, 4, ...
## $ added_at                           <chr> "1970-01-01T00:00:00Z", "19...
## $ is_local                           <lgl> FALSE, FALSE, FALSE, FALSE,...
## $ primary_color                      <lgl> NA, NA, NA, NA, NA, NA, NA,...
## $ added_by.href                      <chr> "https://api.spotify.com/v1...
## $ added_by.id                        <chr> "", "", "", "", "", "", "",...
## $ added_by.type                      <chr> "user", "user", "user", "us...
## $ added_by.uri                       <chr> "spotify:user:", "spotify:u...
## $ added_by.external_urls.spotify     <chr> "https://open.spotify.com/u...
## $ track.artists                      <list> [<data.frame[2 x 6]>, <dat...
## $ track.available_markets            <list> [<"AD", "AE", "AR", "AT", ...
## $ track.disc_number                  <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
## $ track.duration_ms                  <int> 154251, 196583, 269002, 185...
## $ track.episode                      <lgl> FALSE, FALSE, FALSE, FALSE,...
## $ track.explicit                     <lgl> TRUE, FALSE, FALSE, TRUE, F...
## $ track.href                         <chr> "https://api.spotify.com/v1...
## $ track.is_local                     <lgl> FALSE, FALSE, FALSE, FALSE,...
## $ track.name                         <chr> "Wir sind Kral", "AYA", "To...
## $ track.popularity                   <int> 61, 70, 74, 78, 81, 77, 60,...
## $ track.preview_url                  <chr> NA, NA, "https://p.scdn.co/...
## $ track.track                        <lgl> TRUE, TRUE, TRUE, TRUE, TRU...
## $ track.track_number                 <int> 11, 1, 1, 1, 1, 1, 6, 1, 1,...
## $ track.type                         <chr> "track", "track", "track", ...
## $ track.uri                          <chr> "spotify:track:2KlbLTnQ5Wch...
## $ track.album.album_type             <chr> "album", "single", "single"...
## $ track.album.artists                <list> [<data.frame[2 x 6]>, <dat...
## $ track.album.available_markets      <list> [<"AD", "AE", "AR", "AT", ...
## $ track.album.href                   <chr> "https://api.spotify.com/v1...
## $ track.album.id                     <chr> "3gSlFOtOhV1OS3qTQx4g55", "...
## $ track.album.images                 <list> [<data.frame[3 x 3]>, <dat...
## $ track.album.name                   <chr> "Lights Out", "AYA", "Sarka...
## $ track.album.release_date           <chr> "2019-11-14", "2019-09-20",...
## $ track.album.release_date_precision <chr> "day", "day", "day", "day",...
## $ track.album.total_tracks           <int> 12, 1, 5, 1, 1, 1, 12, 1, 1...
## $ track.album.type                   <chr> "album", "album", "album", ...
## $ track.album.uri                    <chr> "spotify:album:3gSlFOtOhV1O...
## $ track.album.external_urls.spotify  <chr> "https://open.spotify.com/a...
## $ track.external_ids.isrc            <chr> "DECE71900765", "NLG6619008...
## $ track.external_urls.spotify        <chr> "https://open.spotify.com/t...
## $ video_thumbnail.url                <lgl> NA, NA, NA, NA, NA, NA, NA,...
## $ key_name                           <chr> "A", "F", "G", "E", "F#", "...
## $ mode_name                          <chr> "major", "minor", "major", ...
## $ key_mode                           <chr> "A major", "F minor", "G ma...
## $ track_sentiment                    <fct> Turbulent/Angry, Happy/Joyf...
## $ artist_names                       <fct> Ufo361, Murda, Sagopa Kajme...

5.1. Country Playlists by Key

The number of common songs in the 4 lists we have is few (Analysis 5.2). In this respect, the musical keys used mainly in songs on a country basis can be obtained from the plot.

country_by_key <- combined_lists%>%
  select(playlist_name, key_name, track.name)%>%
  group_by(playlist_name) %>% count(key_name, sort = TRUE)
  
kable(country_by_key) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed", "responsive"))

ggplot(country_by_key, aes(x = key_name, y = n, fill = playlist_name)) + 
  geom_bar(stat = "identity") +
  labs(title = "Playlists by Key Name", x = "Key Name", y = "Total Number of Keys") + 
  theme(title = element_text(size = 16, face = "bold"), plot.title = element_text(hjust = 0.5), 
        axis.title.x = element_text(size = 14, face = "bold"), 
        axis.title.y = element_text(size = 14, face = "bold"), legend.title = element_blank())

5.2. Common Songs in Playlists

It is possible that there are common songs that could affect our analysis using the music charts of the 4 countries we have. The following table and plot show the songs that are common in the charts and how many of these songs are on the playlists.

common_songs <- combined_lists %>% group_by(track.name, artist_names) %>%
  summarise(n_songs = n()) %>% 
  filter(n_songs >= 2) %>% 
  arrange(desc(n_songs))
kable(common_songs) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed", "responsive"))

ggplot(common_songs, aes(x = reorder(track.name, n_songs), y = n_songs, fill = artist_names)) + 
  geom_bar(stat = "identity") + 
  labs(title = "Common Songs on Playlists", x = "Song Name", y = "Number of Songs") + 
  theme(title = element_text(size = 16, face = "bold"), plot.title = element_text(hjust = 0.5), 
        axis.title.x = element_text(size = 14, face = "bold"), 
        axis.title.y = element_text(size = 14, face = "bold"), legend.title = element_blank()) +
  coord_flip()

5.3 .Danceability Density of Playlists

Danceability shows us how fun and full of energy a song is. So if we can plot the distribution of danceability by country, we can see which country’s playlist is more fun and full of energy. But we need to wait a little bit for the following analysis to detect the emotions and feelings that the songs reflect.

ggplot(combined_lists, aes(x = danceability, fill = playlist_name)) + 
  geom_density(alpha = 0.7, color = NA)+
  labs(x = "Danceability", y = "Density") +
  guides(fill = guide_legend(title = "Playlist"))+
  theme_minimal()+
  ggtitle("Distribution of Danceability Data") +
  theme(title = element_text(size = 16, face = "bold"), plot.title = element_text(hjust = 0.5), 
        axis.title.x = element_text(size = 14, face = "bold"), 
        axis.title.y = element_text(size = 14, face = "bold"), legend.title = element_blank())

5.4. Energy and Valence Range of Playlists

In the steps above, we added a column of sentiment to the songs. We used energy and valence values to do this. The following chart shows the range of energy and valence values according to the country playlists.

playlist_feature_range <- combined_lists %>%
  group_by(playlist_name)%>%
  mutate(max_energy=max(energy), max_valence = max(valence))%>%
  mutate(min_energy=min(energy), min_valence = min(valence))%>%
  select(playlist_name, min_energy, max_energy, min_valence, max_valence)%>%
  unique()
kable(playlist_feature_range) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed", "responsive"))

5.4.1. Energy and Valence Range of Playlists with Dumbbell Plot

The following chart is created with plotly, so you can move your cursor over the chart to see the maximum and minimum values.

energy_range_plot <- plot_ly(playlist_feature_range, color = I("gray80"),  
                hoverinfo = 'text') %>%
  add_segments(x = ~max_energy, xend = ~min_energy, y = ~playlist_name, yend = ~playlist_name, showlegend = FALSE) %>%
  add_segments(x = ~max_valence, xend = ~min_valence, y = ~playlist_name, yend = ~playlist_name, showlegend = FALSE) %>%
  add_markers(x = ~max_energy, y = ~playlist_name, name = "Maximum Energy Value", color = I("red"), size = 2.5, text=~paste('Max Energy: ', max_energy)) %>%
  add_markers(x = ~min_energy, y = ~playlist_name, name = "Minimum Energy Value", color = I("blue"), size = 2.5, text=~paste('Min Energy: ', min_energy))%>%
  add_markers(x = ~max_valence, y = ~playlist_name, name = "Maximum Valence Value", color = I("#395B74"), size = 2.5, text=~paste('Max Valence: ', max_valence)) %>%
  add_markers(x = ~min_valence, y = ~playlist_name, name = "Minimum Valence Value", color = I("#F7BC08"), size = 2.5, text=~paste('Min Valence: ', min_valence))%>%
  layout(
    title = "Playlist Energy and Valence Range",
    xaxis = list(title = "Energy and Valence"),
    yaxis= list(title="Country Lists"))
ggplotly(energy_range_plot)

5.5. Excitement of Playlists

While researching for Spotify analysis, we came across a beautiful analysis on the link. We have updated the formula used in this analysis to suit our own purpose. Here, we imagine what kind of music will be fun to hear when we go to a party or a festival. Values such as energy, danceability, tempo and loudness have a great impact on the energy and happiness of a song. We add the valence value that we will use in the emotion analysis of the songs to this group and use the following formula to find the average excitement of playlists.

excitement_of_playlist <- combined_lists %>% group_by(playlist_name) %>% 
  select(playlist_name, track.name, valence, energy, loudness, danceability, tempo) %>% 
  mutate(excitement = loudness + tempo + (energy*100) + (danceability*100) + (valence*100), excitement_mean = mean(excitement))
ggplot(excitement_of_playlist, aes(x = excitement, fill = playlist_name, color = playlist_name)) + 
  geom_histogram(binwidth = 30, position = "identity", alpha = 0.7) +
  geom_vline(data = excitement_of_playlist, aes(xintercept = excitement_mean, color = playlist_name),
             linetype = "dashed") +
  labs(title = "Excitement Distribution of Playlists", y = "Count", x = "Excitement Scale") +
  theme(title = element_text(size = 16, face = "bold"), plot.title = element_text(hjust = 0.5), 
        axis.title.x = element_text(size = 14, face = "bold"), axis.title.y = element_text(size = 14, face = "bold"),
        legend.title = element_blank())

5.6. Mean Excitement of Playlists

The plot below shows the mean excitement of the playlists by country. Brazil’s playlist is noticeably exciting and entertaining, with Japan in second place out of three other countries whose averages are close to each other.

excitement_mean <- excitement_of_playlist %>% group_by(playlist_name) %>% select(excitement_mean) %>% unique()
kable(excitement_mean) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed", "responsive"))

ggplot(excitement_mean, aes(x = reorder(playlist_name, excitement_mean), y = excitement_mean, fill = playlist_name)) + 
  geom_bar(stat ="identity") + 
  labs(title = "Excitement Comparison of Playlists", x = "Country Playlist Names", y = "Means of Excitement", fill = "Country Charts", 
       caption = "The low score shows that the list is boring. \n Excitement Formula: (loudness + tempo + (energy*100) + (danceability*100) + (valence*100))") +
  theme(title = element_text(size = 16, face = "bold"), plot.title = element_text(hjust = 0.5), 
        axis.title.x = element_text(size = 14, face = "bold"), 
        axis.title.y = element_text(size = 14, face = "bold"), 
        legend.title = element_blank())

5.7. Sentiment Analysis of Country Playlists

We create the a table by selecting energy, valence and sentiment columns for each country.

sentiment_by_countries <- combined_lists %>% group_by(playlist_name) %>% 
  select(playlist_name, track.name, artist_names, valence, energy, track_sentiment)
kable(head(sentiment_by_countries,n=15L)) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed", "responsive"))

5.7.1. Sentiment Analysis of Country Playlists with Gradient Chart

The following table shows the sentiment intensity of the songs in the playlists, grouped by country. The analysis shows that Brazilian playlist is in the Happy/Joyful sentiment class. In other playlists, although the Happy/Joyful is superior, the Turbulent/Angry appears to be in close numbers.

kable(sentiment_by_countries %>% count(track_sentiment, sort = TRUE)) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed", "responsive"))

ggplot(sentiment_by_countries,aes(x = valence, y = energy, color = track_sentiment)) + geom_point() +
  labs(color = "", title = "Sentiment Analysis by Each Country") +
  theme(title = element_text(size = 16, face = "bold"), plot.title = element_text(hjust = 0.5), 
        axis.title.x = element_text(size = 14, face = "bold"), 
        axis.title.y = element_text(size = 14, face = "bold"), legend.title = element_blank()) +
  scale_x_continuous(expand = c(0, 0), limits = c(0, 1)) + 
  scale_y_continuous(expand = c(0, 0), limits = c(0, 1)) +
  geom_label(aes(x = 0.12, y = 0.98, label = "Turbulent/Angry"), label.padding = unit(1, "mm"), fill = "grey", color="white") +
  geom_label(aes(x = 0.90, y = 0.98, label = "Happy/Joyful"), label.padding = unit(1, "mm"), fill = "grey", color="white") +
  geom_label(aes(x = 0.12, y = 0.025, label = "Sad/Depressing"), label.padding = unit(1, "mm"),  fill = "grey", color="white") +
  geom_label(aes(x = 0.895, y = 0.025, label = "Chill/Peaceful"), label.padding = unit(1, "mm"), fill = "grey", color="white") +
  geom_segment(aes(x = 1, y = 0, xend = 1, yend = 1)) +
  geom_segment(aes(x = 0, y = 0, xend = 0, yend = 1)) +
  geom_segment(aes(x = 0, y = 0, xend = 1, yend = 0)) +
  geom_segment(aes(x = 0, y = 0.5, xend = 1, yend = 0.5)) +
  geom_segment(aes(x = 0.5, y = 0, xend = 0.5, yend = 1)) +
  geom_segment(aes(x = 0, y = 1, xend = 1, yend = 1)) +
  facet_wrap(~ playlist_name)

6.1. Monthly Change in Total Streams

Firstly, we wanted to look at monthly total stream change. In the graph below, we see that total stream increases exponentially. We can conclude that usage of spotify and stream amount increased rapidly in Turkey.

#First cal
#topturkey200 %>% group_by(Artist)%>% summarise(Total_number=n()) %>% arrange(desc(Total_number))
change<-topturkey200 %>% mutate(Year_Month = format(Date,"%Y/%m")) %>% group_by(Year_Month) %>% summarise(Total_Stream=sum(as.numeric(Streams))) 
ggplot(change, aes(x = Year_Month,y=Total_Stream,group=1)) + geom_point() + geom_smooth() + theme(axis.text.x = element_text(angle = 90),title = element_text(size = 16, face = "bold"), plot.title = element_text(hjust = 0.5)) + labs(x = "Month", y = "Total Streams",title = "Total Stream Change") + scale_y_continuous(labels = comma)

6.2. Most Streamed 20 Tracks

If we curious about which songs streamed most between 2017 and 2019 so far, Ezhel - Geceler is the answer. First 20 most streamed songs are displayed in the table below.

rank<-topturkey200 %>% group_by(Artist,Track.Name) %>% summarise(Total_Stream=sum(as.numeric(Streams))) %>% arrange(desc(Total_Stream))
kable(head(rank,n=20L)) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed", "responsive"))

6.3. Sentiment Analysis of Tracks

6.3.1. Data Preparation

For sentiment analysis, we need to extract unique songs in data frame. So that, we can determine the sentiments of each tracks which can entry the top 200 songs in sometime. After this process, we extract the Spotify ID’s of tracks on a new column for further analysis.

top_200<-topturkey200 %>% mutate(id=substring(topturkey200$URL,32)) 
top_200_unique<-top_200[!duplicated(top_200[,c('id')]),]
glimpse(top_200_unique)

## Observations: 2,874
## Variables: 7
## $ Position   <chr> "1", "2", "3", "4", "5", "6", "7", "8", "9", "10", ...
## $ Track.Name <chr> "Gece Gölgenin Rahatina Bak", "Starboy", "Affet", "...
## $ Artist     <chr> "Çagatay Akman", "The Weeknd", "Müslüm Gürses", "Cl...
## $ Streams    <chr> "80607", "44427", "34889", "28400", "25425", "23032...
## $ URL        <chr> "https://open.spotify.com/track/3P31rcl0ym5paqRdwSi...
## $ Date       <date> 2017-01-01, 2017-01-01, 2017-01-01, 2017-01-01, 20...
## $ id         <chr> "3P31rcl0ym5paqRdwSiZps", "5aAx2yezTd8zXrkmtKl66Z",...

With track ID’s, we obtanied audio features of this songs using Spotify Web API and “spotifyr” R package. To increase process speed, we upload the RDS data in our github repository and read it from there.

Id_list=top_200$id
#Tracks feautres obtaining code is below. Because of the process time data frame downloaded from github repository.
#a<-unique(Id_list)
#tracks_features=get_track_audio_features(a[1])
#for (x in 2:length(a)){
#  tracks_features <- rbind(tracks_features,get_track_audio_features(a[x]))
#}
#tracks_features<-tracks_features%>%slice(-1) 
 
tracks_features<-readRDS(url("https://github.com/pjournal/mef03g-spo-R-ify/blob/master/top200_tracks_features.rds?raw=true"))
glimpse(tracks_features)

## Observations: 2,874
## Variables: 18
## $ danceability     <dbl> 0.769, 0.681, 0.424, 0.720, 0.476, 0.748, 0.4...
## $ energy           <dbl> 0.837, 0.594, 0.666, 0.763, 0.718, 0.524, 0.3...
## $ key              <int> 6, 7, 7, 9, 8, 8, 4, 4, 1, 5, 6, 6, 0, 8, 7, ...
## $ loudness         <dbl> -4.057, -7.028, -6.683, -4.068, -5.309, -5.59...
## $ mode             <int> 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 0, 0, 1, 0, 1, ...
## $ speechiness      <dbl> 0.1400, 0.2820, 0.0473, 0.0523, 0.0576, 0.033...
## $ acousticness     <dbl> 0.65200, 0.16500, 0.39600, 0.40600, 0.07840, ...
## $ instrumentalness <dbl> 0.00e+00, 3.49e-06, 1.69e-04, 0.00e+00, 1.02e...
## $ liveness         <dbl> 0.0986, 0.1340, 0.1200, 0.1800, 0.1220, 0.111...
## $ valence          <dbl> 0.8190, 0.5350, 0.2750, 0.7420, 0.1420, 0.661...
## $ tempo            <dbl> 100.962, 186.054, 160.079, 101.965, 199.864, ...
## $ type             <chr> "audio_features", "audio_features", "audio_fe...
## $ id               <chr> "3P31rcl0ym5paqRdwSiZps", "5aAx2yezTd8zXrkmtK...
## $ uri              <chr> "spotify:track:3P31rcl0ym5paqRdwSiZps", "spot...
## $ track_href       <chr> "https://api.spotify.com/v1/tracks/3P31rcl0ym...
## $ analysis_url     <chr> "https://api.spotify.com/v1/audio-analysis/3P...
## $ duration_ms      <int> 163960, 230453, 279475, 251088, 205947, 24496...
## $ time_signature   <int> 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, ...

After we prepared 2 data frame, we joined them in to one data frame. Than, we add a column in our new data frame using our sentiment function.

top_200_audio_features <- inner_join(top_200_unique,tracks_features,by="id")
Sentiment=c()
for (i in 1:nrow(top_200_audio_features)){
  Sentiment[i]=classify_track_sentiment(valence=top_200_audio_features$valence[i],energy=top_200_audio_features$energy[i])
}
top_200_audio_features<-cbind(top_200_audio_features,Sentiment)
glimpse(top_200_audio_features)

## Observations: 2,874
## Variables: 25
## $ Position         <chr> "1", "2", "3", "4", "5", "6", "7", "8", "9", ...
## $ Track.Name       <chr> "Gece Gölgenin Rahatina Bak", "Starboy", "Aff...
## $ Artist           <chr> "Çagatay Akman", "The Weeknd", "Müslüm Gürses...
## $ Streams          <chr> "80607", "44427", "34889", "28400", "25425", ...
## $ URL              <chr> "https://open.spotify.com/track/3P31rcl0ym5pa...
## $ Date             <date> 2017-01-01, 2017-01-01, 2017-01-01, 2017-01-...
## $ id               <chr> "3P31rcl0ym5paqRdwSiZps", "5aAx2yezTd8zXrkmtK...
## $ danceability     <dbl> 0.769, 0.681, 0.424, 0.720, 0.476, 0.748, 0.4...
## $ energy           <dbl> 0.837, 0.594, 0.666, 0.763, 0.718, 0.524, 0.3...
## $ key              <int> 6, 7, 7, 9, 8, 8, 4, 4, 1, 5, 6, 6, 0, 8, 7, ...
## $ loudness         <dbl> -4.057, -7.028, -6.683, -4.068, -5.309, -5.59...
## $ mode             <int> 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 0, 0, 1, 0, 1, ...
## $ speechiness      <dbl> 0.1400, 0.2820, 0.0473, 0.0523, 0.0576, 0.033...
## $ acousticness     <dbl> 0.65200, 0.16500, 0.39600, 0.40600, 0.07840, ...
## $ instrumentalness <dbl> 0.00e+00, 3.49e-06, 1.69e-04, 0.00e+00, 1.02e...
## $ liveness         <dbl> 0.0986, 0.1340, 0.1200, 0.1800, 0.1220, 0.111...
## $ valence          <dbl> 0.8190, 0.5350, 0.2750, 0.7420, 0.1420, 0.661...
## $ tempo            <dbl> 100.962, 186.054, 160.079, 101.965, 199.864, ...
## $ type             <chr> "audio_features", "audio_features", "audio_fe...
## $ uri              <chr> "spotify:track:3P31rcl0ym5paqRdwSiZps", "spot...
## $ track_href       <chr> "https://api.spotify.com/v1/tracks/3P31rcl0ym...
## $ analysis_url     <chr> "https://api.spotify.com/v1/audio-analysis/3P...
## $ duration_ms      <int> 163960, 230453, 279475, 251088, 205947, 24496...
## $ time_signature   <int> 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, ...
## $ Sentiment        <fct> Happy/Joyful, Happy/Joyful, Turbulent/Angry, ...

kable(head(top_200_audio_features,n=10L)) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed", "responsive")) %>% scroll_box(width = "100%", height = "400px")

6.3.2. Sentiment Change in Months

For proper sentiment analyse, we calculate the percentages of sentiments frequency in every month. With percentages, we can compare the users emotional preferences in months.

df1a<-top_200 %>% mutate(Year_Month = format(Date,"%Y/%m"))
df1b<-df1a %>% left_join(select(top_200_audio_features, "Sentiment","id"), by = "id")
monthly_sentiment<-df1b %>% group_by(Year_Month,Sentiment) %>% summarise(Count_Sentiment = n()) %>% ungroup() %>% group_by(Year_Month) %>% mutate (Month_Sum=sum(Count_Sentiment)) %>% ungroup() %>% mutate(Percent_in_Month = percent(Count_Sentiment/Month_Sum))
glimpse(monthly_sentiment)

## Observations: 140
## Variables: 5
## $ Year_Month       <chr> "2017/01", "2017/01", "2017/01", "2017/01", "...
## $ Sentiment        <fct> Chill/Peaceful, Happy/Joyful, Sad/Depressing,...
## $ Count_Sentiment  <int> 186, 2710, 848, 2456, 151, 2452, 690, 2307, 1...
## $ Month_Sum        <int> 6200, 6200, 6200, 6200, 5600, 5600, 5600, 560...
## $ Percent_in_Month <chr> "3.0%", "43.7%", "13.7%", "39.6%", "2.7%", "4...

kable(head(monthly_sentiment,n=10L)) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed", "responsive"))

monthly_sentiment<-monthly_sentiment%>% mutate(Perc_Num=as.double( strsplit(Percent_in_Month,split = "%")))
ggplot(monthly_sentiment,aes(Year_Month,Perc_Num,group=Sentiment,color=Sentiment)) + geom_point() + geom_line(aes( color=Sentiment)) +  labs(title = "Monthly Sentiment Change", x = "Months", y = "Percantage") +
  theme(title = element_text(size = 16, face = "bold"), plot.title = element_text(hjust = 0.5),axis.text.x = element_text(angle = 90), 
        axis.title.x = element_text(size = 14, face = "bold"), 
        axis.title.y = element_text(size = 14, face = "bold"), legend.title = element_blank()) 

6.3.3. Sentiment Change in Years

Spotify offer us Turkey’s playlists in 1980, 1990, 2000, 2010 decades. These playlists includes just 50 songs per decade. Thus, we use sentiment percentage frequency again for compare decades and last 3 years data.

df80<-get_playlist_audio_features("spotifycharts","37i9dQZF1DX4io1yPyoLtv") %>% mutate(Year="1980") 
df90<-get_playlist_audio_features("spotifycharts","37i9dQZF1DXb7MJRXLczzR") %>% mutate(Year="1990") 
df00<-get_playlist_audio_features("spotifycharts","37i9dQZF1DWYteTcNVQZNq") %>% mutate(Year="2000") 
df10<-get_playlist_audio_features("spotifycharts","37i9dQZF1DXaE9T4Nls8eC") %>% mutate(Year="2010") 
past_track_data <- rbind(df80, df90, df00, df10)
glimpse(past_track_data)

## Observations: 200
## Variables: 62
## $ playlist_id                        <chr> "37i9dQZF1DX4io1yPyoLtv", "...
## $ playlist_name                      <chr> "Türkçe 80'ler", "Türkçe 80...
## $ playlist_img                       <chr> "https://pl.scdn.co/images/...
## $ playlist_owner_name                <chr> "Spotify", "Spotify", "Spot...
## $ playlist_owner_id                  <chr> "spotify", "spotify", "spot...
## $ danceability                       <dbl> 0.395, 0.742, 0.469, 0.534,...
## $ energy                             <dbl> 0.675, 0.186, 0.486, 0.739,...
## $ key                                <int> 1, 8, 9, 4, 2, 4, 9, 7, 7, ...
## $ loudness                           <dbl> -5.881, -16.820, -15.076, -...
## $ mode                               <int> 0, 1, 0, 1, 0, 0, 0, 0, 1, ...
## $ speechiness                        <dbl> 0.0472, 0.0449, 0.0427, 0.1...
## $ acousticness                       <dbl> 0.6490, 0.8220, 0.1490, 0.6...
## $ instrumentalness                   <dbl> 0.00e+00, 0.00e+00, 1.21e-0...
## $ liveness                           <dbl> 0.0963, 0.1620, 0.1230, 0.2...
## $ valence                            <dbl> 0.496, 0.570, 0.595, 0.505,...
## $ tempo                              <dbl> 177.053, 114.982, 130.737, ...
## $ track.id                           <chr> "2rP7pI2WpMWcUraYAX2xiT", "...
## $ analysis_url                       <chr> "https://api.spotify.com/v1...
## $ time_signature                     <int> 4, 4, 4, 4, 3, 4, 5, 4, 4, ...
## $ added_at                           <chr> "2019-08-07T12:49:23Z", "20...
## $ is_local                           <lgl> FALSE, FALSE, FALSE, FALSE,...
## $ primary_color                      <lgl> NA, NA, NA, NA, NA, NA, NA,...
## $ added_by.href                      <chr> "https://api.spotify.com/v1...
## $ added_by.id                        <chr> "", "", "", "", "", "", "",...
## $ added_by.type                      <chr> "user", "user", "user", "us...
## $ added_by.uri                       <chr> "spotify:user:", "spotify:u...
## $ added_by.external_urls.spotify     <chr> "https://open.spotify.com/u...
## $ track.artists                      <list> [<data.frame[1 x 6]>, <dat...
## $ track.available_markets            <list> [<"AD", "AE", "AR", "AT", ...
## $ track.disc_number                  <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
## $ track.duration_ms                  <int> 259360, 257044, 239266, 229...
## $ track.episode                      <lgl> FALSE, FALSE, FALSE, FALSE,...
## $ track.explicit                     <lgl> FALSE, FALSE, FALSE, FALSE,...
## $ track.href                         <chr> "https://api.spotify.com/v1...
## $ track.is_local                     <lgl> FALSE, FALSE, FALSE, FALSE,...
## $ track.name                         <chr> "Tükenecegiz", "Bu Kalp Sen...
## $ track.popularity                   <int> 60, 48, 55, 50, 35, 56, 49,...
## $ track.preview_url                  <chr> "https://p.scdn.co/mp3-prev...
## $ track.track                        <lgl> TRUE, TRUE, TRUE, TRUE, TRU...
## $ track.track_number                 <int> 8, 1, 1, 6, 2, 2, 7, 8, 6, ...
## $ track.type                         <chr> "track", "track", "track", ...
## $ track.uri                          <chr> "spotify:track:2rP7pI2WpMWc...
## $ track.album.album_type             <chr> "album", "album", "album", ...
## $ track.album.artists                <list> [<data.frame[1 x 6]>, <dat...
## $ track.album.available_markets      <list> [<"AD", "AE", "AR", "AT", ...
## $ track.album.href                   <chr> "https://api.spotify.com/v1...
## $ track.album.id                     <chr> "13JKU1RyLFS73hDvWnTHr1", "...
## $ track.album.images                 <list> [<data.frame[3 x 3]>, <dat...
## $ track.album.name                   <chr> "Sen Aglama", "Bu Kalp Seni...
## $ track.album.release_date           <chr> "1984-09-06", "2002-02-20",...
## $ track.album.release_date_precision <chr> "day", "day", "day", "year"...
## $ track.album.total_tracks           <int> 11, 9, 10, 10, 10, 13, 11, ...
## $ track.album.type                   <chr> "album", "album", "album", ...
## $ track.album.uri                    <chr> "spotify:album:13JKU1RyLFS7...
## $ track.album.external_urls.spotify  <chr> "https://open.spotify.com/a...
## $ track.external_ids.isrc            <chr> "TR0061200300", "TR00806002...
## $ track.external_urls.spotify        <chr> "https://open.spotify.com/t...
## $ video_thumbnail.url                <lgl> NA, NA, NA, NA, NA, NA, NA,...
## $ key_name                           <chr> "C#", "G#", "A", "E", "D", ...
## $ mode_name                          <chr> "minor", "major", "minor", ...
## $ key_mode                           <chr> "C# minor", "G# major", "A ...
## $ Year                               <chr> "1980", "1980", "1980", "19...

Sentiment=c()
for (i in 1:nrow(past_track_data)){
    Sentiment[i]=classify_track_sentiment(valence=past_track_data$valence[i],energy=past_track_data$energy[i])
}
past_track_data<-cbind(past_track_data,Sentiment) 
past_track_data_sentiment<-past_track_data %>% group_by(Year,Sentiment) %>% summarise(Count= n())
df1c<-top_200 %>% mutate(Year = format(Date,"%Y"))
sent_count_yearly<-df1c %>% left_join(select(top_200_audio_features, "Sentiment","id"), by = "id") %>% group_by(Year,Sentiment) %>% summarise(Count= n())
yearly_change<- rbind(past_track_data_sentiment,sent_count_yearly) %>% group_by(Year) %>% mutate (Year_Sum=sum(Count)) %>% ungroup() %>% mutate(Percent_in_Year = percent(Count/Year_Sum))
kable(yearly_change) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed", "responsive"))

yearly_change<-yearly_change%>% mutate(Perc_Num=as.double( strsplit(Percent_in_Year,split = "%")))
ggplot(yearly_change,aes(Year,Perc_Num,group=Sentiment,color=Sentiment)) + geom_point() + geom_line(aes( color=Sentiment)) +  labs(title = "Yearly Sentiment Change", x = "Years", y = "Percantage") +
  theme(title = element_text(size = 16, face = "bold"), plot.title = element_text(hjust = 0.5),axis.text.x = element_text(angle = 90), 
        axis.title.x = element_text(size = 14, face = "bold"), 
        axis.title.y = element_text(size = 14, face = "bold"), legend.title = element_blank()) 

6.3.4. Sentiment Bar Graph of All Tracks

Sentiment count of all unique songs between 2017 and 2019.

sent_count <- top_200_audio_features %>% group_by(Sentiment) %>% count()
ggplot(sent_count, aes(x=Sentiment, y=n, fill=Sentiment)) +
  geom_bar(stat="identity") + 
  labs(title = "Sentiment Count", x = "Sentiment Distribution", y = "Count of Sentiments") +
  theme(title = element_text(size = 16, face = "bold"), plot.title = element_text(hjust = 0.5), 
        axis.title.x = element_text(size = 14, face = "bold"), 
        axis.title.y = element_text(size = 14, face = "bold"), legend.title = element_blank())

6.3.5. Sentiment Gradient Chart

Finaly we mapped the all Top 200 tracks by their sentiment and displayed in gradient chart.

ggplot(top_200_audio_features,aes(x = valence, y = energy, color = Sentiment)) + geom_point() +
  labs(color = "", title = "Sentiment Analysis of Turkey Top 200 Chart Between 2017 and 2019") +
    theme(title = element_text(size = 16, face = "bold"), plot.title = element_text(hjust = 0.5), 
        axis.title.x = element_text(size = 14, face = "bold"), 
        axis.title.y = element_text(size = 14, face = "bold"), legend.title = element_blank()) +
  scale_x_continuous(expand = c(0, 0), limits = c(0, 1)) + 
  scale_y_continuous(expand = c(0, 0), limits = c(0, 1)) +
  geom_label(aes(x = 0.25, y = 0.97, label = "Turbulent/Angry"), label.padding = unit(2, "mm"),  fill = "darkgrey", color="white") +
  geom_label(aes(x = 0.75, y = 0.97, label = "Happy/Joyful"), label.padding = unit(2, "mm"), fill = "darkgrey", color="white") +
  geom_label(aes(x = 0.25, y = 0.03, label = "Sad/Depressing"), label.padding = unit(2, "mm"),  fill = "darkgrey", color="white") +
  geom_label(aes(x = 0.75, y = 0.03, label = "Chill/Peaceful"), label.padding = unit(2, "mm"), fill = "darkgrey", color="white") +
  geom_segment(aes(x = 1, y = 0, xend = 1, yend = 1)) +
  geom_segment(aes(x = 0, y = 0, xend = 0, yend = 1)) +
  geom_segment(aes(x = 0, y = 0, xend = 1, yend = 0)) +
  geom_segment(aes(x = 0, y = 0.5, xend = 1, yend = 0.5)) +
  geom_segment(aes(x = 0.5, y = 0, xend = 0.5, yend = 1)) +
  geom_segment(aes(x = 0, y = 1, xend = 1, yend = 1)) 

7.1. Audio Features Analysis by Radar Chart

Click on the link to use our app, which analyzes and compares the audio features of music charts created by Spotify or belonging to two different Spotify users as a radar chart.

7.2. Musical Horoscope

Click on the link to use our app, which makes predictions on personality type based on the user playlist, using the audio features and key characteristic of the Spotify user’s playlist.