Content-Based Recommendation of Songs from Spotify

Source: Photo by Alexander Shatov on Unsplash

Project Summary

In this project I am implementing a content-based recommendation system using a K-Means clustering algorithm to predict songs (pieces) from Spotify based on a playlist from a user.

The resulting K-Means cluster looks like this:

These clusters are then used for the content-based recommendation.

How can you use it?

Technical Requirements

• Python 3.8
• Pandas
• scikit-learn
• matplotlib
• seaborn
• scipy

Files

The important files are:

• music_recommender.ipynb: analysis of data from Spotify and building a content-based recommender for tracks
• blog_post.md: Demystifying Recommendations By Predicting Your Next Favorite Songs (Or Pieces) - an evaluation of the data from Spotify and Introduction to the Music Recommender
• data: the data I used can be found under data/

Credits and Acknowledgments

Must give credit to Yamac Eren Ay who created the dataset as part of a Kaggle competition and inspired my work. You can find the data and competition here. Additionally, I would like to thank Spotify for providing a Web API and access to the features associated with the tracks. Feel free to use the code here as you would like!