charliehyin / musicrecommender

A Jupyter Notebook that predicts song popularity and performs k-means clustering.

MusicRecommender

Utilizes a dataset of 25000+ songs and their features to perform song clustering and popularity prediction

Clustering is done using K-means clustering and DBScan clustering.

Popularity prediction is done with a decision tree classifier. The classifier can predict a song's popularity with about 70% accuracy.

Deatures used by the decision tree classifier:

• danceability
• energy
• key
• loudness
• speechiness
• acousticness
• liveness
• valence
• tempo
• duration_ms

Features in the data:

• Artist Name
• Track Name
• Popularity
• Genres
• Playlist
• danceability
• energy
• key
• loudness
• mode
• speechiness
• acousticness
• instrumentalness
• liveness
• valence
• tempo
• id
• uri
• track_href
• analysis_url
• duration_ms
• time_signature