Ana Boccanfuso
Music plays a significant role in all of our lives: connecting us with others, allowing us to relate to one another, and eliciting emotions out of us. And yet, music is also something that makes us all unique. As each person has a music taste that is distinct like a fingerprint, music streaming platforms have implemented data analysis to try to pinpoint an individual’s music taste and predict songs they may enjoy. These algorithms are particularly useful in discovering new music, and we can learn more about our personal music taste by analyzing what we listen to and how often.
There is so much to explore within music. Some of these include variations within specific genres, variations within an artist’s discography, attributes of popular songs, and the evolution of music trends over time. Exploring these categories will allow for a better understanding of an “audio identity” for specific music taste, allowing listeners to have a better relationship with their music library.
To begin analyzing music, we must look at the history of music. For this project, pop music is being studied. The following visualizations depict the evolution of pop music from 1960 to 2015, illustrating how specfic attributes have changed over time.
The first line graph depicting annual change in average pop elements (danceability, valence, energy) shows some fascinating trends over the years. First, the average valence, which is the overall positiveness of a song, has decreased by about 0.12 points since 1960. Additionally, valence rose significantly during the 1980s, then decreased relatively rapidly until 2000. To understand this, we must consider the music popular during this period. The 1980s experienced music from Michael Jackson, Prince, Madonna, and Whitney Houston. As some of the most iconic figures in pop music, this decade saw an emergence of dance music, which is supported by the increase in average danceability. This graph also illustrates music’s increase in average energy since 1960. Energy reached a peak in 2011 at 0.75, 0.3 greater than its value in 1960. The peak in energy can be explained by the rise in EDM (electronic dance music) around this time.
The second line graph looks at speechiness, acousticness, and liveness. Liveness refers to how many tracks appear to be performed live. This value remains relatively steady from 1960 to 2015. Speechiness measures the presence of spoken words in a track. This value also appears to stay around the same, with an increase in the early 2000s. The increase may be justified by pop-rap artists such as Eminem and Kanye. The most dramatic change occurred in acousticness, which decreased from 0.65 (1960) to 0.17 (2015). As pop music has evolved, more technologies have been used such as synthesizers and mixers. This digitalization of music creation has therefore decreased the acousticness of pop music over the years.
The next element of pop music that was explored was explicitness. As the graph shows, the Billboard Top 100 was never explicit between 1960 and 1982. Explicitness began significantly increasing in the late 90s, with a peak in 2005 at 0.34. The drastic increase in explicitness can be explained by the rise in rap and artists such as Britney Spears.
We now perform a case study on the artist that has the most monthly listeners on Spotify, Taylor Swift. Recently, Taylor Swift has broken countless record with her new album Midnights. She became the first artist in history to occupy the entire top 10 of the Billboard Hot 100, Midnights became Spotify’s most-streamed album in a single day, she broke the record for the most-streamed artist in a single day in Spotify history, and many others.
Below, I have visualized the valence and danceability of Swift's discography. These plots take the individual valence/danceability values of each song within an album, producing a holistic view of an album's valence/danceability distribution shape and range. The curve of a density plot allows inference of how Swift's albums are balanced relative to each other, as well as her evolution over time.
In the last part of my analysis, I wanted to compare how much I listen to music throughout the year, as well as how my favorite artists change based on time I spend listening to them and how many distinct tracks I listen to. I also wanted to visualize who my top artist is throughout the year, as it may change. This data was collected by requesting my personal streaming history from Spotify. Below, I have visualized these different aspects in several types of charts. You can toggle what's selected by choosing an artist.
Focusing on the first scatter plot, it is evident that the two artists that I listen to the most are The 1975 and Taylor Swift. I have played both of these artists the most over the past year, with total minutes played averaging 7,400 minutes. However, I have listened to more distinct tracks by Taylor Swift than The 1975, explained by Taylor’s 10 albums compared to The 1975’s five albums. This graph also shows how much more I listen to these two artists than others combined. My next most listened to artist is The Lumineers, with 1,400 minutes played and 60 distinct tracks. Therefore, I disproportionately listen to my two favorite artists in comparison, in both listening time and distinct tracks listened to. This finding is further supported by the heat map on the right. Although the size of the boxes show how much I favor my two favorite artists, the shading also provides great insight. The 1975 and Taylor Swift are a deep green, and the next shaded artists (The Lumineers and Harry Styles) have a very light shade of green in comparison. The other artists in the heat map have a color that barely resembles green.
For the next graph in my dashboard, I plotted my most listened to artist throughout the year, using a running sum of the minutes played. Taylor Swift is consistent throughout the year until the very end, where The 1975 barely surpassed her. This shows a shift in my listening pattern at the end of the year when I started to strictly listen to the band. The growth of Taylor’s curve slowed as The 1975’s curve shot up. Looking at the end of the graph, we again see the disparity between total minutes played of my top two artists and my next three top artists. Each of these artist’s curves did not grow much in comparison. For example, looking at Harry Styles’ curve, I listened to him for about one month before the curve flattened (indicating I stopped listening to his music). Khalid’s flat curve was more consistent but much lower.
For the last visualization in my dashboard, I created a line graph depicting how much music I listen to each day over the past year. I found it interesting how contrasting this line was, with some days reaching hundreds of minutes of listening, and others amounting to a mere four minutes. However, it is evident that I listened to more music later in the year, starting in June of 2022.
All of the data used in this project was gathered through requested data from the Spotify API and two R packages: spotifyr and billboard.
Using the billboard r package, I accessed a data frame containing the top 100 Billboard hits in the United States between 1960 and 2015. Within these years, I analyzed the average music elements for each year including danceability, energy, valence, speechiness, acousticness, and liveness. Each of these measures have a value between 0.0 and 1.0, and are determined by Spotify.
