Spotify-Clustering

Live 3d-plot

An ongoing effort to leverage the Spotify API and perform various tasks like :-

• gathering the user's top tracks along with audio features and saving it as a csv
• gathering the user's top artists and saving it as a csv
• creating a playlist of top tracks
• getting all liked songs, creating feature vectors and clustering them in 3d (using t-SNE and k-means)

Setting up

Install the required modules with $ pip install -r requirements.txt

To actually call the spotify api, you would need to create a demo app on developer.spotify.com to get a CLIENT_ID and CLIENT_SECRET for your app.

Now create a .env file with your CLIENT_ID and CLIENT_SECRET tokens like :-

CLIENT_ID=*******************
CLIENT_SECRET=****************

Running the script

Gathering top tracks and saving to disk as csv. These tracks can be either long term(default), short term or medium term specified with the -t flag. Filename is specified by the -n flag. Use -h or --help for further details.

python main.py top_tracks -t long -n long_term_fav_tracks.csv

Gathering top artists and saving to disk as csv These artists can be either long term, short term or medium term specified with the -t flag. Use -h or --help for further details

python main.py top_artists -t long -n long_term_fav_artists.csv

Creating a playlist of top tracks. Time range is again specified with the -t flag.

python main.py playlist -t long

Clustering all liked songs (more tracks are usually needed for clustering) Number of clusters is specified with the -k flag, 3 by default.

python main.py cluster -k 3

Below is a link to a live 3d plot of all my liked songs clustered with k=3
Live 3d plot