Looking for clarity on the no ML/AI developer policy for a personal project

[aj_hoyt3]

I am a rising Junior majoring in data science in the U.S. For a personal project this summer, I thought it would be interesting to develop a recommendation system for the student radio at my university that tries to cater to an indie/alt crowd. Ideally the app uses the web API to access a users playlist, take in the data from the songs about the playlist, and then would use the recommendation endpoint to recommend a playlist with similar songs to the given playlist that are under the popularity threshold the station uses to that of the songs in a playlist (< 10M total streams, <1M artist monthly listeners). Of course I have to estimate the streams and monthly listeners with the popularity index because that data is not available through the API, nor anywhere else programmatically for free. 

To find similar songs, I figured I would use some sort of clustering algorithm that would identify the similarities in the songs audio features to find general target values for input into the recommendation system. However I remembered the developer policy that says "Do not use the Spotify Platform or any Spotify Content to train a machine learning or AI model or otherwise ingest Spotify Content into a machine learning or AI model." 

Would this fall under the above policy? I do not intend to deploy this app to production or monetize it in any way, I only to use this as a project to showcase and further develop my skills as a programmer. 

Is anyone able to provide clarity on this issue, or know of someone to reach out to? 

In the meantime, I'd be happy to hear any alternatives to ML that are still good ways to find songs with similar attributes to generalize and use as input into the recommendation endpoint. 

[Ximzend]

You could better use the existing endpoint: Get Recommendations.

[Ximzend]

To answer your question, the Developer Terms and Developer Policy apply to all endpoints, no matter what kind of project you use them for.

[aj_hoyt3]

Can you explain please? the best way i can think of without breaking ML/AI policy is to use the song ids as seeds and set a max popularity value arbitrarily or letting the user pick.

[Ximzend]

Yes, exactly that.

[CRobbo]

Did you go about developing this idea? I'm looking to do something for a Uni project that will use recommendations and unsure whether it breaks any TOS. But if worst comes to worst, i can always just create the recommendations from prechosen seeds, that the model knows about. So i guess that would be skirting around the TOS without breaking any rules