hectorfa17/clustering-songs

Song Recommender

This repository is used to recommend songs from an input name

Overview

I created a song recommender based on the features of the songs. It does the modeling of the data and assigns a cluster number to each song. It is also able to give a new cluster number to each new song. From there it can recommend other songs also depending on the 100-hot-songs list. It follows the schematic that can be found on the end of this document.

You can find the app on this link

Contents

• Notebooks

1. src/1.get_song_lists.ipynb - Used to get the songs lists and find the song features for each list
2. src/2.clustering_songs.ipynb - Clustered the songs using 2 different models: K-means and BDSCAN
3. src/3.song_recomender.ipynb - Created the song recommender step by step
4. src/4.song_recomender_function.ipynb - Created the function "user friendly"

• Libraries

1. src/spoty_jzar.py - This is the class that is used as a Spotify interface
2. src/music_jzar.py - this lib is used to scrap and get the raw lists
3. src/cluster_jzar.py - this lib is used for clustering functions

Databases

• I used a scrapper to get the 100 hot songs from
• For the big dataset I used https://www.kaggle.com/datasets/rodolfofigueroa/spotify-12m-songs

Requirements

• beautifulsoup4==4.10.0
• matplotlib==3.3.4
• numpy==1.19.5
• pandas==1.4.1
• requests==2.25.1
• scikit_learn==1.0.2
• scipy==1.6.0
• seaborn==0.11.2
• spotipy==2.19.0
• streamlit==1.8.0

Schema

This is the schematic of the song recommender