🎧 Spotify Music Discovery

Create your own Spotify recommendation algorithm. All you need is Python 3, Jupyter Notebook, and a Spotify account.

Setup

1. Clone/download the repository.
2. Ensure that you have Python 3 and Jupyter Notebook installed.
3. Navigate to /spotify-music-discovery and run pip install -r requirements.txt.
4. Create the file spotify-music-discovery/pkg/config.py as below, using your Spotify client information.

CLIENT_ID = 'your Spotify client id here'
CLIENT_SECRET = 'your Spotify client secret here'
CLIENT_USERNAME = 'your Spotify username here'

5. To help avoid any OAuth errors that might occur, open your Spotify application here, and set up your Redirect URIs as follows:

• Edit Settings
  • Add http://localhost:8888/callback/ under "Redirect URIs"

6. When prompted in get_spotify_data, copy and paste the link you are redirected to in the input box that should pop up after running the first cell (even if the link throws a "localhost redirected you too many times" error).

Usage

• Start Jupyter Notebook in the /spotify-music-discovery directory.
• Run through the notebooks in sequence, following the instructions in each:
  1. get_spotify_data
     • Downloads and parses song data from the training playlists you specify.
  2. train
     • Trains classifiers using the training data and saves the best one.
       • Specifically, it pickles the classifier object and writes it to the ./classifiers directory.
  3. predict
     • Predicts which songs you like from a playlist you specify.

Spotify URIs

To download playlists, you will have to specify their Spotify URIs. You can get Spotify URIs from the Spotify app, as follows:

• Right-click on a playlist
  • Share
    • Copy Spotify URI

Background

This repo was originally created for a final project in an applied machine learning course at the University of Pennsylvania. For more detail, including the algorithms used, see the project report.