Do variations in audio attributes such as loudness, valence etc. have substantial influence on the popularity of songs on Spotify?
In the era of music streaming, the factors that contribute to the popularity of certain tracks have become a topic of interest in recent times. The central question guiding this analysis is ‘Do variations in audio attributes such as loudness, valence etc. have substantial influence on the popularity of songs on Spotify?’. In this report we uncover the relationships between the popularity of songs on Spotify and audio attributes that a particular track has. Spotify, being one of the most widely used platforms in the world today with over 551 million monthly active users, led us to choose the dataset related to the platform. The predictors of interest in our analysis include both continuous and categorical variables. Continuous predictors, such as danceability, energy, tempo, and loudness, offer insights into the quantitative aspects of a track's composition. On the other hand, categorical predictors, including time signature and mode, contribute to our understanding of the qualitative musical attributes that might impact popularity. It should also be noted that ‘musical key,’ though treated as a categorical variable, is transformed into a dummy variable so that we can accurately grasp the nuances that are associated with different musical keys. In our study we used a linear regression model to perform the data analysis because it allows us to express the relationship between the chosen predictor variables and popularity. The coefficients of the respective predictor variables will provide us with an intuition of their impact on the popularity of the track. In the quest of understanding the factors that influence track popularity on Spotify, it is crucial to first examine and understand existing knowledge pertaining to the research question. The article “Music Popularity: Metrics, Characteristics, and Audio-Based Prediction” aims to measure the popularity of a track through sales and the number of streams. However, there is a gap in knowledge on how certain aspects of the track itself determine popularity. The article “What Makes a Music Track Popular in Online Social Networks?” correlates popularity with music content, the artist’s reputation, and the social context of the track. Like the first article, the authors used predictor variables that measure external factors rather than the musical aspect of the tracks. The article “A Model for Predicting Pop Music Popularity and Its Different Characteristics Based on Multiple Linear Regression” measured the popularity of pop music on YouTube through streams and musical-related predictors that we plan to use. However, the gap in knowledge is that the article does not cover many genres of music. Also, the article uses YouTube as the music platform instead of Spotify. This can impact results because the demographics of users differ between platforms, and some songs might be exclusive to a certain platform. These peer-reviewed articles provide us existing knowledge and helps us identify gaps, to provide a more intricate understanding of the research question. Nevertheless, most articles aimed to predict the popularity metrics of the study. We aim to restrict ourselves into finding the most appropriate and best model to find the associations between popularity and other covariates.