Enhancing Spotify Playlists Using Their Audio Features With Classical And Deep Learning Methods

We used the Spotify Web API to extract the data, which allowed us to access the audio features of every song and the contents of thousands of playlists.

Using the data we gathered, we developed two models that could generate recommendations. The first model was a K-Nearest Neighbors (KNN) clustering algorithm. This algorithm works by using the audio features of a song to create clusters of songs that are similar to each other. It then uses a user’s preferences to recommend songs from the clusters they have not heard before. The second model was a deep learning approach that used a neural network. This model works by using the audio features of songs to train the neural network to accurately predict a user’s preferences.