Build a Spotify-Like Music Recommender System in Python

Learn how to build a music recommender system that suggests music artists using collaborative filtering and Alternating Least Squares. In the process, you’ll get a foundation about music recommenders, how Spotify recommends music, and how to use a production-ready collaborative filtering Python library called ‘implicit’.

======================================

Get the code:

======================================

The Sound of AI Academy:

The Sound of AI Recruitment:

Advanced Python Programming:

Join The Sound Of AI Slack community:

Interested in hiring me as a consultant/freelancer?

Connect with Valerio on Linkedin:

Follow Valerio on Facebook:

Follow Valerio on Twitter:

======================================

Content

0:00 Intro
0:28 Types of music recommendation
2:17 Spotify recommendation architecture
3:41 What are we building?
4:18 User-item matrix
6:41 Alternating least squares
16:46 Package overview
17:28 Data loader routine
20:30 Artist retriever
23:12 Implicit - a collaborative filtering library
23:47 Recommender implementation
27:55 Running artist recommendations
30:35 Join The Sound of AI Slack Community!

[ValerioVelardoTheSoundofAI]

This video scratches the surface of music recommendation. Let me know if you'd be interested in a full series / live workshop on the topic by leaving a comment 👇

[zzzz]

Wow Valerio great video! I just finished a final project in this topic!

[manamahani]

even though 2 years have passed I wanted to thank you. this video helped me lots!

[YelongHe]

Hi Valerio! Loved your video, and thanks for a easy understandable explanation on how a recommendation system works. I was wondering if there were a newer version of the lastFM dataset, than 2011?

[christian_sigmas-ckyv]

can you show how this recommendation works with flutter firebase project?

[doyourealise]

hey, amazing! ive watching your video after a long time. Hope you are doing well.

[matimg]

Hi Valerio,

I wanted to express my gratitude for taking the time to create this incredibly informative and well-explained video! It's been a tremendous help.

However, I couldn't help but notice what appears to be a minor error in the way you implemented the `recommend` method. In your implementation, you utilized the `n` attribute as an index for the `user_artist_matrix` Matrix. However, according to Implicit's official documentation, it should be the `user_id`, and `n` should ideally be its independent argument, denoted as "N"

Here's the snippet from the official documentation for reference:

"""
# Get recommendations for the a single user
userid = 12345
ids, scores = model.recommend(userid, user_plays[userid], N=10,
"""

Hope this helps clarify things, and again, big thanks for your fantastic work!

Also, let me know if I'm the one misunderstanding something!

[rohanbobhate]

That video is really quite helpful. Can we do it by extracting data from youtube personal account and predict prefered content ?

[AdrianFernandezFazio]

Great video Valerio! In the first minutes of the video there's a map of the Spotify 's architecture. What's the purpose of the CNN in Discover Weekly?

[Msyoutube]

Hey got a question, can you elaborate what user and artist matrix columns and rows represent?

[aravindsbabu]

is this model use collabarative filtering algorithm?

[isururanaweera]

How can we use this to recommend artists to a new external user??

[wanderingtravellerAB]

Have always loved your content Valerio, thanks for posting.

[itssahilonyoutube]

can the same code be implemented for c++? please reply author . thanks again for the idea!!

[HuyTao-]

It said No module named 'musiccollaborativefiltering' even though i already run poetry install, please help me, im in the middle of my school project