Having collected millions of concerts via the SongKick API in my database, and having analysed the profiles of each artist via the Spotify API, I had the idea to create FestivalPlaylist.io.
The principle is simple: you connect to your Spotify account, and each week FestivalPlaylist.io generates a personalised playlist from your favourite festival for you to view.
How does the algorithm work?
It's very basic.
The user chooses a festival with a line-up they like
The user connects to their Spotify account and authorises access to it
The tool analyses the user's favourite artists and creates a typical profile of the user's "tastes"
Then, it selects artists corresponding to the user's profile from the line-up, and intersperses this with artists the user is already listening to (the algorithm produces a 50/50 mix of the two)
Afterwards it saves the generated Playlist on the user's account
Every week until the festival takes place, a new playlist will be generated for the user.
This method isn't perfect: the tool will offer users the artists they already like, or artists in the genre they usually listen to, but it won't offer any artists they might like in a genre they don't know yet.
This is where the difficulty in writing recommendation algorithms lies. That said, the fact that the line-up for the festival was (largely) the work of one person already gives the selection of artists the human curation element we're looking for here when creating recommendations.
One idea to overcome the problem could be, for example, to choose artists at random from the line-up and drag them into a generated Playlist consisting 90% of artists we're sure the user likes, or may like.
Want to try? Go to festivalplaylist.io