MelodyMap

A graph-based music recommendation system that finds similar songs based on audio features and creates personalized Spotify playlists.

Originally developed as a collaborative project, then extended independently with additional optimizations and features.

Overview

This application analyzes Spotify song data to find tracks with similar audio characteristics. When you select a song, the system:

1. Calculates similarity scores against thousands of songs using audio features
2. Returns the top 25 most similar tracks with similarity percentages
3. Automatically creates a Spotify playlist with your recommendations
4. Visualizes the song similarity network as an interactive graph

Features

• Smart Song Search: Search by song name or artist with auto-suggestions
• Audio-Based Matching: Uses 11 audio features for accurate similarity scoring
• Spotify Integration: Creates public playlists directly in your Spotify account
• Network Visualization: Interactive graph showing song relationships
• Multi-Song Mode: Find songs similar to two tracks simultaneously

How It Works

Similarity Algorithm

Audio features are normalized using MinMaxScaler, then songs are compared using weighted Euclidean distance across 11 Spotify audio features. The result is converted to a similarity score between 0% (completely different) and 100% (identical features).

Feature	Description
Danceability	How suitable for dancing (0.0 - 1.0)
Energy	Intensity and activity level (0.0 - 1.0)
Valence	Musical positiveness/happiness (0.0 - 1.0)
Tempo	Beats per minute (BPM)
Acousticness	Confidence the track is acoustic
Speechiness	Presence of spoken words
Instrumentalness	Likelihood of no vocals
Liveness	Probability of live recording
Loudness	Overall volume in decibels
Key	Musical key (0-11)
Mode	Major (1) or minor (0)

Graph Structure

• Each song is a vertex containing metadata and audio features
• Edges connect songs with similarity scores above 90%
• The graph enables efficient traversal and relationship visualization

Installation

Prerequisites

• Python 3.9+
• Spotify Developer Account (for API access)

Setup

1. Clone the repository:

   git clone https://github.com/oliverbisbee/MelodyMap
   cd MelodyMap

2. Install dependencies:

   pip install spotipy networkx matplotlib scikit-learn

3. Configure Spotify API credentials in main.py:

   SPOTIFY_CLIENT_ID = "your_client_id"
   SPOTIFY_CLIENT_SECRET = "your_client_secret"
   SPOTIFY_REDIRECT_URI = "http://127.0.0.1:8080"

4. Run the application:

   python main.py

Usage

Single Song Mode (main.py)

1. Launch the GUI with python main.py
2. Type a song name or artist in the search box
3. Select a song from the suggestions
4. Click "Select Song" to get recommendations
5. View similarity scores and the generated playlist link
6. Double-click the playlist link to open in Spotify

Multi-Song Mode (multiple song.py)

1. Run python "multiple song.py"
2. Enter two song names when prompted
3. View songs that are similar to both tracks
4. See the dual-similarity visualization

Project Structure

MelodyMap/
├── main.py                 # Main application with GUI
├── multiple song.py        # Two-song similarity comparison
├── ex3_visualization.py    # Alternate visualization module
├── spotify_songs.csv       # Dataset (~32,000 songs)
└── spotify_data/
    └── readme.md           # Dataset documentation

Dataset

The dataset contains approximately 32,000 Spotify songs with:

• Track metadata (name, artist, album)
• Audio features extracted via Spotify API
• Popularity scores

Source: TidyTuesday Spotify Dataset

Visualization

The network graph displays:

• Red node: Your selected song
• Green nodes: Similar songs (90%+ similarity)
• Edges: Connections between highly similar tracks

Dependencies

• spotipy - Spotify API wrapper
• scikit-learn - Feature normalization and similarity computation
• networkx - Graph data structures
• matplotlib - Visualization
• tkinter - GUI framework (included with Python)

Authors

Oliver Bisbee, Mustafa Maj, Rayyan Hussain, Mahir Asif Chowdhury