Just Landed...

Last week we had country - level finals of the Microsoft Imagine Cup competition and again I was one of the judges. Our work starts much earlier, as in April we have to select 10 teams that will be competing during May finals. When I was working through this year's set of submitted projects, my first impression was 2009 students show less of the "imagination factor" than they used to in previous years. However, when the general level seems to be lower, there are still some diamonds hidden and our job (as a jury) is to find them.

I have been especially impressed by one project, submitted by the Nosoi Fighters team. What they do is complex network modeling, simulation and prognosis of various factors. Based on the underlying knowledge and technology they presented the CARE application. The primary goal of the CARE is to predict how various diseases are going to spread, and then to estimate vaccination strategy, including distribution of vaccines, how much and where will be needed and so on.

I have never been before to network modeling, but I really like the idea. Of course the theoretical model of a social network is one thing and real world connections between people are something different. And the real difficulty of bringing the model as close to reality as possible is to gather enough real world data. While one could crawl Facebook for network of connections, this would not get us anywhere. Facebook represents virtual connections. I may have many friends there I have never met in person. And for viral diseases we still (fortunately!) have to be physically at least in a proximity to infected people to catch an infection. Twitter's "follow me" model does not help her either.

But both Facebook and Twitter have something else we may use as an input to the model. Statuses. What are you doing now? "Just landed in Chicago..." Are you getting the message? Analyzing status updates we can start building a model of people traveling across the globe. Both Facebook and Twitter have home town properties. A point to start. And then following the posted status updates we may reflect the way people move in geo coordinates. I started digging around the idea and found this simulation: http://vimeo.com/4587178 by Jer Thorp. Based on the very analysis of Twitter livestream filtered to start with "Just landed" Jer proved he could implement a proof of concept model of air travel. Surely this is just a starting point, as for certain scenarios we may look for "I feel sick" or similar sentences or better build a semantic model trying to understand the meanings of status updates.

As I wrote before, Twitter (and it's value) is not understood by many. But there is a reason Google and others offer sums in excess of a 1 Billion dollars for a platform that lets users enter 140 - character micro messages. To me the value of course is in Twitter having reached critical mass with its popularity skyrocketing in recent weeks / months. But it is also in the idea of making everything entered there public. Nobody will ever legally read your emails, but with Twitter you agree you have no rights to whatever you put in... So the information can be then propagated to various third party systems thriving on it.

Complex networks modeling of course has many other applications. Epidemics is just one area, but similar models and techniques can be applied in marketing or fault detection / propagation or many other areas. Bartosz Lipinski of the Nosoi Fighters promised me they would run a simulation based on Twitter's data. I will be checking their web site.