Ever wondered why the suggestions on shopping websites such as amazon or ebay more often than not appeal to you? Most links are targeted and do related to recent activity linking activity and interest together.
Amazon's recommendations for products especially books works as a network of relations and starts to groups together similar interest areas. This is based on cross user activity and behaviour with a learning environment. One purchase leads to another and so on.
Christocper Warnow has looked int o creating a network visualisation for the amazon recommendation service and has written a Processing app making use of the open source Gephi API. The tool can take a web link to a book on amazon and create a network around it for up to 100 recommendations associated with the publication.
The tool can be downloaded HERE for a test. It rus in real time and the process of building the network is unfolding on screen, quite interesting to follow. The tool allows for zooming in/out and hovering for information as well as an pdf export function of the created network visualisation.
Image by Christopher Warnow / Recommendation network on amazon.de for Linked
Image by Christopher Warnow / Recommendation network on amazon.com for Linked
Interesting are the differences between the amazon online stores. Warnow points out that there are connection recommendation differences between for example the .de and the .com store in some examples: "I wanted to compare the recommendations given by amazon.de and amazon.com. And another surprise waited. The milieus looked different. The Germans are connecting postmodernism with Deleuze, the buyers from amazon.com are thinking more about the French situationist movement. I tested it again with the awesome book Linked by Albert-Laszlo Barabasi. The similarities are the complex theory and network dynamics. But the differences are interesting as well. The English milieus contain politics and collective systems, where as the Germans are more into successful marketing and economics."