There is a part of my Data Mining course I am very fond of. This is Data Visualization, a matter often overlooked when implementing DM processes and one -and this is my contention-  of extreme relevance to the whole enterprise.  The human visual system is an extremely subtle and beautiful natural pattern recognition engine that, to our benefit, can be seamlessly embedded within the human-crafted pattern recognition techniques. What follows aims to be an ad hoc repository of resources on data visualization. It is also meant to be a showcase of my students' contribution to this part of the course.

Have a look at these examples of excellence in data visualization provided by this year's students (as presented in the seminar 10/10/07). They are just a sample of how smart visualization can get in this Internet age:

1. When visualization rings you a bell

One way to hit the bull’s eye in data visualization is by grounding it in a directly unrelated way of visualization that, for whatever the reason, is deeply ingrained in human visual reasoning. Chernoff’s faces are an example of that, in which data attributes are coded as facial features, expanding (at least a bit) the possibilities of low-dimensional representations of multivariate data.

A really elaborated example, provided by Javi Dolcet, is this representation of web trends for 2007 created by IA (Information Architects, Japan), in which the underlying representation is Tokio’s metro map. Most of us, urbanites, already know how to interpret a map such as this, despite its complexity. A fully interactive version of this map can be found here

... And a truly beautiful meta-example, provided by Álvaro García, in which a palette of visualization methods are themselves visualized using, as underlying representation, Mendeleev's periodic table (which, by itself, is a great visualization example). A fully interactive web version of the table can be found here.

2. Visualizing relationships

Guillermo Nebot provided an example of visualization of social networks (by eye-sys, but given that the company seems to have discontinued the product, we are not including the image here). There is plenty of work on this type of visualization, some of which is compiled here. As an alternative example, have a look at this representation of the interactions between characters in the novel Les Miserables by Victor Hugo, divided in communities represented by different colors  (by M. Newman and M. Girvan, at University of Michigan, Santa Fe Institute, Cornell University; USA)

3. Geographies

Cartograms are a nice way to use geographical information as the basis of data representation. They can be defined as maps in which areas are distorted according to a quantitative, area-related variable, and in which map contiguity contrains are preserved. The following example, representing HIV prevalence by nation, was presented by Ángela Chieppa, and can be found here.