You are not logged in.
Here is my problem:
I'm trying to develop a way to rank music files. Now, all of the music files include metadata on the artist, trackname, length, etc...
Say I have some partial data taken from last.fm. This includes most listened to tracks, artists, as well as banned tracks, and favourited tracks.
This does NOT include all of the music files.
So the set of tracks with data is a subset of all of the tracks.
let
D = the set of all tracks with data
A = the set of all of the tracks
The problem is how I go about ranking the files (say from 1 to n).
I could simply go through and mark each track that I know has data on it, maybe 3 for a track that is favourited, 2 for a most listened to track, 1 for a track with no data on it, and 0 for a banned track.
However D is a very small subset of A.
How do I go about ranking A based purely on D?
Are there some basic techniques, perhaps based on statistics or probability that I can use? My first thought is something similar to google's page rank. Maybe I first check if a file is in D, and if not, I go to the last.fm page of a random artist in D, then I proceed to check if my file's artist is linked to from that page. If not I go to each of the artists listed on the page (under the similar artists) and see if it's linked to from there, keeping track of the number of levels of recursion. Maybe I do this for a set number of levels (to avoid it becoming exponential). Then I would simply take the number of levels it took, and give the file that number as rank.
Does this sound like a good start? Any other good ideas? Would this be too inefficient for large A?
Please be gentle with the math. I'm not extremely knowledgeable.
Last edited by Nisstyre56 (2011-10-07 01:23:03)
In Zen they say: If something is boring after two minutes, try it for four. If still boring, try it for eight, sixteen, thirty-two, and so on. Eventually one discovers that it's not boring at all but very interesting.
~ John Cage
Offline