• 1 Introduction
• 2 Getting the data
• 3 Analysing and Visualizing the data
• 4 Analysis
• 4.1 Most collaborations
• 4.1 Clusters
• 4.2 Key collaborations between clusters
• 5 Wrap up

1 Introduction

As a Hip-hop fan (and other genres of music) it is obvious that certain artists are more alike than others. I was interested to see if a network of Hip-hop artists based on their collaborations with others would be able to highlight some interesting insights.

The things I needed to solve this problem were:

• Data on artists, their tracks (songs) and the artists featuring on the tracks
• Network graph analysis and visualisation software

This was done on a Windows 10 machine with:

• Python 3
• Node.js 16
• Gephi

2 Getting the data

To have a look at the source code for getting the collaboration data, have a look at the following Github repo https://github.com/mortie23/musicbrainz-get. I looked for existing dataset but counldn't find one, but found a fantastic free API from MusicBrainz https://musicbrainz.org/doc/MusicBrainz_API/Examples.

To get started, since it is 2023, I started with a prompt to ChatGPT that was something like this:

Write me a Python script that calls that Music Brainz API (documentaiton here https://musicbrainz.org/doc/MusicBrainz_API/Examples) that searches for an artist and returns the artists ID.

Then, with a bit of refactoring and a few more prompts to ChatGPT, I ended up with a script that could loop through a list of my favourite artists and produce a CSV file of the collaboration graph. The analysis was a bit of a throw back to the 1990's of Hip-hop and I wanted a few East coast and West coast rappers in the list to see if the network analysis clusters them well.

# %%
artist_list = [
    "Jay-Z",
    "Nas",
    "Wu-Tang Clan",
    ...
]

3 Analysing and Visualizing the data

After quickly trying some network and charting packages in Python, and then trying a PowerBI markplace viz with minimal success, I remembered a desktop appliction I used years ago that could render beautiful network graphs. So I started working with Gephi https://gephi.org/users/download/.

4 Analysis

Here is the first look of the network graph with labels. The size of the nodes (and labels) is the number of edges in the graph (the number of collaboration artists). The color is:

• Orange: Calfornia
• Green: New york
• Blue: Chicago

4.1 Most collaborations

The winner of this title goes to Snoop Dogg.

4.1 Clusters

The east coast and west coast artists are clustered well together based on their collaborations (as expected).

4.2 Key collaborations between clusters

It seems from the network graph that Dr Dre (from the west coast) signing 50 Cent (from the east coast) to his label became a key collaboration connection between the clusters.

5 Wrap up

I think there is great insights that can be gained from collaborations between artists that confirm some suspicions of fans and highlight some unknown insights.