This is an old project that I recently replicated in Python. This study is focused on my Facebook friends data, collected through Facebook Graph API. I collected my list of friends and the list of music pages each of them liked.

I treated the json data to structure the data in a matrix format with shape n X m, where n is the number of user friends and m is the number of music pages. I considered only the 80 most liked pages. As illustrated in following:

Them I first applied an SVD technique, to create a reduced number (12) of latent variables (dimensions), where I could note some similarities between artists of these pages.

At the end, I applied a clustering technique using these 12 dimensions, where I could find clusters of friends who likes more specific genres of music, e.g. cluster of friends who like Rock Music most, or cluster of friends who like Samba and Pagode the most.

The Colab Notebook of this study is available here.

The data sets are available here.

• Singular Value Decomposition
• Hierarchical Clustering
• K-Means
• Music
• Facebook
• Social Network