Skip to content

Latest commit

 

History

History
30 lines (24 loc) · 1.33 KB

README.md

File metadata and controls

30 lines (24 loc) · 1.33 KB
Raw

The-Bjorker

Are you as inspired as Bjork?

Using machine learning and Spotipy, this app compares the songs on Bjork's favorite albums to a playlist of your choice. The songs will be determined as inspiring to Bjork or not quite. The results will surprise you, so feel free to keep trying different playlists!

Data pipeline

  • Extracts songs data from Spotify with Redis queued workers using API wrapper Python package spotipy with pandas based on inputs and sets related to this project (i.e. Bjork's top 10 inspiring albums dataset part of training and validation set)
  • Transforms track analysis data into usable tables, with values rescaled and used to train and test a machine learning model with scikit and keras
  • Loads downstream data after applying predict functions with model, visualizing test data results in Flask app deployed on Heroku

Monitoring & Logs

DataDog

Future Features

  • Use URLs as well as URIs as input
  • Able to enter multiple playlists at once
  • Create joint playlists by inspiring tracks
  • Create merged track lists by non-Bjork-approved songs
  • More dashboard style options to visualize tracks data
  • More!

To Run

  • python wsgi.py to start the local server
  • localhost:5000 to reach in your browser

App currently deployed on heroku at http://bjorker.herokuapp.com