"Data Visualization at Microsoft"

4/28/2000: [time not recorded]
[location not recorded]

Abstract: With large amounts of data accumulating in databases and data warehouses, there is a growing need for tools that help find interesting patterns from data. One general class of such tools goes by the name "data visualization." In my talk, I will discuss data visualization techniques that have been developed at Microsoft. These tools have been used internally and will likely appear in future data analysis products. The tools are somewhat different from classic methods for data visualization which typically display raw data. The tools I will discuss first abstract the data using statistical models and then present the models to users in a (visual) way that facilitates the discovery of patterns. I will highlight visualization techniques based on two models: dependency networks -- a close cousin of the Bayesian network -- and mixture models. I will illustrate the use of these tools using data from domains including web navigation, ecommerce, TV viewing, and movie preferences.

Last updated: Mon Jun 19 17:44:06 2006