I have always been interested in music. I grew up playing several instruments and took a hiatus from university to pursue music production in a professional manner. Unfortunately that pursuit did not work out, but I still make music as a hobby. Because of this interest, while scrolling through kaggle a few months ago, one specific dataset caught my eye, the Spotify dataset. Throughout my journey in music creatively and professionally I was always curious about the trends I saw in music across time and across popularity. Using this dataset we will attempt to satisfy some of my curiosity about the properties of popular music.

Because this dataset is incredibly large, we will narrow down the scope of our exploration slightly with specific questions that we will attempt to answer to prevent scope creep. These questions are:

• How has music as a whole changed over time?

• How has popular music changed over time?

• Are there any specific traits of a song that would indicate greater popularity? Can we create a model that predicts the popularity of a song based on its properties?

We will use a host of different strategies to answer these questions, mostly consisting of machine learning and visualization.

Quick and dirty.