Show HN: Using Social Network Analysis to predict your event's guest list | Hacker News Reader

Show HN: Using Social Network Analysis to predict your event's guest list

whodidiforget.com

16 pointsfriggeri11 comments

Given that my work as a PhD student focuses on social groups, I thought I'd launch a little website to showcase the applications of what I do. I'd appreciate any feedback you might have.

For those interested in the gory little details, all the analysis is done client side, in Javascript (as a matter of fact, the whole app is just a static website hosted on github pages). If some are interested, I might blog about the different aspects of the implementation.

I usually will not connect anything to my facebook, but just curious: does it establish these relationships based on my interaction and networks shared with my friends, or some other method?

The only information it uses is "who is friends with who". I'm not looking at messages/interactions/networks/existing groups/friends list. Just the underlying graph of friendships.

Please do blog about the different aspects of implementation. Or at least post the link to the Github repo. This seems very interesting, and I'd love to hear more about how you did it.

Unfortunately, the github repo is private because I can't open source the whole algorithm right now. But I'll definitely do a write up in the next few days.

I blogged about the science behind the app here : http://friggeri.net/blog/who-did-i-forget

This is fascinating! Has it been accurate?

> I'm not looking at messages/interactions/networks/existing groups/friends list. Just the underlying graph of friendships.

Other interactions may give you better signals may increase the accuracy of your system. Cool app!

Of course, that's be ideal, but it would take longer to compute the weights. The idea here was really to show that just by looking at friendships it's possible to find accurate results.

There are a lot of apps in Facebook that does weight the friendships and generated images. Maybe one of them have their algorithm opened and you may use it to improve the accuracy.

I'll rephrase that: adding a weight to an edge is not hard, I actually having several heuristics on hand to do that. It's just really time consuming the first time you do it, because you have to load a lot of data in order to be accurate. Moreover it requires additional permissions, which I'd rather avoid.

Reply on news.ycombinator.com