recommendation engine based on playcount, need data
If you use the playcounts from the media library from a considerable number of people it should be quite easy to create a recommendation engine using item based recommendation algorithms. I'd like to make this into a winamp plugin. It will send all the playcount and song data to a central server but before it can make any recommendations I'd have to have a considerable amount of data before it would be able to do anything.
Would it be acceptable for the first bunch of users to be stuck with badly fitting recommendations or none at all before the data reaches a critical mass? (it could possibly be 100+)
Do you guys have an idea on how to collect enough of that data?
If anyone feels like donating their data please pm me. Perhaps it will be effective with only a few samples.