Jaidev Deshpande writes:
The World Economic Forum recently posed a data visualization problem. The dataset is a survey of experts from the so called “Agenda Councils” of the WEF. Here are the details.
The dataset primarily contains the experts’ opinions on which global / regional / industrial agenda council of the WEF they would benefit from by interacting with the most. It occurs to me that this dataset can be thought of as an instance of a social networking dynamics, in that it represents the preferences of individuals towards belonging or not belonging to a particular group within the network. It is these ‘groups’ that must be identified to solve the problem. Under what conditions would this hypothesis be valid?
I have a hunch that dimensionality reduction will not necessarily help me visualize this data satisfactorily. They also need to be complemented by the way social networks detect cliques amongst their members.
The prize is $3000 plus bragging rights, and submissions are due 15 Nov.
The given similarities between councils are highly non-metric: in particular, they violate the triangle inequality and are asymmetric. Hence, standard multidimensional scaling techniques will not be able to appropriately visualize the data.
I created an interactive dashboard to filter on various fields and see the graph update. However, I cannot submit this to the contest as a URL: http://visualizefree.com/share.jsp?id=INnIrBNn
It seems that submissions must be hosted by visualizing.org, but the Visualize Free website is served dynamically. There is no way to get a standalone version.