audiofyle

spotify recommendation system using K-NN algorithm and Spotify Web API

origins

While fiddling around with the Spotify Web API, I thought it would be interesting to see how the recommendation algorithm changed when prioritizing different factors, like popularity, genre, tempo, and energy. I began this project thinking that creating a custom deployment to experiment with a basic yet efficient classifying ML algorithm would be a great way to grasp the utility of the model.

stack

• Backend: Flask

  • spotipy: python lib for working with spotify
  • scikit-learn: using scikit-learn for normalized mutual info to test robustness of model

• Frontend: React

  • next.js