Back to Portfolio

Song Recommender

How does it work?

This system uses Machine Learning techniques to find similar songs based on musical features like tempo, danceability, energy and more.

1

Data Preprocessing

We clean the data and use K-means clustering to group similar songs, making recommendations faster and more accurate.

2

KNN Algorithm

We use the K-Nearest Neighbors algorithm to find the most similar songs within each cluster.

3

Similarity Ranking

Songs are ordered by similarity percentage, showing the most similar ones first.

Analyzed Features:

  • Year: Release year of the song
  • Acousticness: How acoustic the song is (0-100%)
  • Artists: Name of the performing artists
  • Danceability: How suitable the song is for dancing (0-100%)
  • Duration: Length of the song in milliseconds
  • Instrumentalness: How instrumental the song is (0-100%)
  • Liveness: Probability of the song being performed live (0-100%)
  • Loudness: Overall volume in decibels (dB)
  • Name: Title of the song
  • Popularity: Song's popularity score (0-100%)
  • Speechiness: Presence of spoken words (0-100%)
  • Tempo: Speed or rhythm in BPM (beats per minute)
View complete project on GitHub

Example Recommendations

Macarena
Los Del Rio
Año: 1996
Popularidad: 59.0%
Duración: 4:09
Volumen: -12.6 dB
Tempo: 103.0 BPM
Bailabilidad: 93.8%

Recommended Songs

Tulsa Time
Don Williams
92.6% similarity
Año: 1978
Tempo: 104.4 BPM
Bailabilidad: 92.7%
I Can't Dance
Genesis
88.3% similarity
Año: 1991
Tempo: 107.6 BPM
Bailabilidad: 92.5%
Pedro Y Pablo
Los Tigres Del Norte
88.3% similarity
Año: 1984
Tempo: 108.9 BPM
Bailabilidad: 87.0%
Nasty
Janet Jackson
87.4% similarity
Año: 1986
Tempo: 103.7 BPM
Bailabilidad: 84.9%
The Rubberband Man
The Spinners
86.3% similarity
Año: 1993
Tempo: 97.3 BPM
Bailabilidad: 82.0%