The Last.FM Similarity Project

Guybrush · 10-09-2009, 03:35 AM

Hey guys!

I thought of the potential for doing statistics with the data gathered from last.fm. One of the easier things one could do is just to check similarity between users and I thought something like that might be interesting to do in our musicbanter community.

The idea is that in the end, we'll get some sort of visual representation - 2d or 3d - in which you can see different people from our community represented as dots or something and the distance between them tells you how similar they are in taste. Something like that is likely to create clusters where members of similar tastes group together and with people whose tastes are a bit broader may be found in between these clusters.

This requires us to make a similarity matrix and contributing to it is going to require a little dedication from all our participants.

Because this will get very labour intensive if there are many people and many people leaving or entering the project, we need to create a list of participants before we start collecting data. I'll set the limit of participants to 20 and to ensure participants are active members of our community, I'll only accept those who have a post count of 100 or more. Also, participants have to have been members of last.fm for a long enough time so that people can get similarity comparisons with them.

Summary : As a participant, first post here that you want to participate. Once all slots are filled (20 people), you'll have to post your similarity scores with the other participants in this thread.

Last.FM Scores :

Very Low - 1
Low - 2
Medium - 3
High - 4
Very High - 5
Super - 6

Participants

Here are participants with links to their last.fm pages. Each participant will be represented in the final "product" with a two-letter (or number) code.

edit :

RESULTS ARE IN

10-09-2009, 03:35 AM	#1 (permalink)
Guybrush Juicious Maximus III Join Date: Nov 2008 Location: Scabb Island Posts: 6,525	The Last.FM Similarity Project Hey guys! I thought of the potential for doing statistics with the data gathered from last.fm. One of the easier things one could do is just to check similarity between users and I thought something like that might be interesting to do in our musicbanter community. The idea is that in the end, we'll get some sort of visual representation - 2d or 3d - in which you can see different people from our community represented as dots or something and the distance between them tells you how similar they are in taste. Something like that is likely to create clusters where members of similar tastes group together and with people whose tastes are a bit broader may be found in between these clusters. This requires us to make a similarity matrix and contributing to it is going to require a little dedication from all our participants. Because this will get very labour intensive if there are many people and many people leaving or entering the project, we need to create a list of participants before we start collecting data. I'll set the limit of participants to 20 and to ensure participants are active members of our community, I'll only accept those who have a post count of 100 or more. Also, participants have to have been members of last.fm for a long enough time so that people can get similarity comparisons with them. Summary : As a participant, first post here that you want to participate. Once all slots are filled (20 people), you'll have to post your similarity scores with the other participants in this thread. Last.FM Scores : Very Low - 1 Low - 2 Medium - 3 High - 4 Very High - 5 Super - 6 Participants Here are participants with links to their last.fm pages. Each participant will be represented in the final "product" with a two-letter (or number) code. +81 - +8 adidasss - AD Anteater - AE anticipation - AP Bardonodude - BA Bulldog - BD Captain Caveman - CC Dreadnaught - DN Jackhammer - JH Kayleigh - KL khfreek - KF LoathsomePete - LP Molecules - MC NumberNineDream - ND Sidewinder - SW storymilo - SM Toretorden - TT Urban Hatemonger - UH wolverinewolfweiselpigeon - WW Zeppy111 - Z1 edit : RESULTS ARE IN __________________ Something Completely Different