Drillable Stacked Time Series

Some random data

About This Visualization

Using Flare's outstanding Job Voyager sample application as a starting point, I extended the stacked time series to support a one-level hierarchy of time series data. Each time series on the initial view is an aggregate of multiple, second level time series. Clicking on any of the time series bands will zoom into the components of that time series, clicking on any of those second level time series returns the view to the first level. I also added a toggle to allow the user to view the data in absolute or normalalized form.

The impetus for the work was the analysis of several years worth of weblog information for a Ruby on Rails web application. There each page served to a user corresponds to an action, and controllers are groups of related actions. It makes sense to view the growth of the site through the growth of each area of the site, as divided up by controller, then zoom in to see which actions on a controller are being used. Are users contributing or viewing content, and does that breakdown change over time?

The Actionscript code for this project is on github. I also wrote an R script to create some fake weblog data, which can be seen here.

About the author

By day, James Kebinger is a Senior Developer at PatientsLikeMe.