A Quantitative Analysis of Spotify Songs

Popularity vs. Individual Audio Features
Danceability vs. Energy vs. Speechiness vs. Popularity
Box Plots of Audio Features Scored From 0 to 1

Inspiration

Music is one thing that can unite people of widely different backgrounds. We used our group's shared love for music to analyze what makes a song popular.

What it does

Our first research question attempts to find trends with audio features and popularity. Our second research question uses Spotify's 'Get Recommendations' endpoint to see if audio features from popular songs will return songs of a similar popularity score. Our final research question attempts to predict a song's popularity score based on its audio features.

How we built it

Our team used a shared GitHub repo to build and collaborate on this project.

Challenges we ran into

One of the main challenges we had was getting our libraries installed correctly. We had issues installing Tensorflow and Keras specifically. Additionally, because Keras was a new library, it was difficult to debug why our ML model was getting an MSE of 4000+ at first.

Accomplishments that we're proud of

We are especially proud of the fact that the ML model we built and tuned ourselves turned out to have the best results.

What we learned

For this project, our group is proud of how we took on learning Plotly and Keras, as well as how to automate our API requests using Python.

What's next for A Quantitative Analysis of Spotify Songs

In the future, we would like to experiment more with finding relationships between specific audio features, and if these relationships can help our ML model and our API request return better results than what we received.