About TreeVersity

TreeVersity explores change in trees over time. Users can use TreVersity to analize relative and absolute change over one variable in each node, as well as created and removed nodes. Moreover, analysts can use TreeVersity to compare non-inherently hierarchichal datasets, by grouping them by it's attributes. For example, users looking at the change on the number of movies watch, can build a hierarchy grouping by age ranges, then by gender and then by favorite music genre, and answer questions like "how has changed the number of terror movies watched by teenage girls in the last year?".

TreeVersity provides a set of tools to facilitate the comparison process that include:

• comparison algorithms that calculate the change over each node of the tree (internal and leaf nodes)
• two visualization techniques called Bullet and Stacked Skylines that represent the change in each node of the tree with controls for navigation and filtering
• and a reporting tool that searches for the most significant changes on the tree and displays them in a textual report linked to the visualizations

Much work has been done on visualizing and exploring single tree structures; however, the problem of comparing two trees is significantly harder. We have identified and classified the following types of tree comparison:

• Type 0: Topological differences between two trees where the nodes only contain a label. Example: Finding created, moved and removed topics between two versions of the TRB research publications hierarchical categorization system, without looking for number of papers.
• Type 1: Positive and negative changes in leaf node values with aggregated values in the interior nodes (i.e. trees that can be visualized with a treemap[25]) and no changes in topology. Example: Comparing the change on the number of workers on the traffic agencies of the country, grouped by Agencies, Counties and Regions, given that no Agencies are created or removed.
• Type 2: Positive and negative changes in leaves and interior node values with no changes in topology. Example: Comparing the salaries in the Department of Transportation’s organizational chart between two years, when no reorganization has occurred.
• Type 3: Positive and negative changes in leaf node values with aggregated values in the interior nodes and with changes in topology. Example: Again finding changes in the TRB research publications hierarchical categorization system, but looking at changes in the number of papers published by topic and at created, removed nodes.
• Type 4: Positive and negative changes in leaves and interior node values, with changes in topology. Example: Comparing the number of TRB’s website visits between two months using the file hierarchy as a natural organization. Some pages might be created or removed, and each page in the hierarchy has an independent number of visits.

Most of the previous work on the field has focused on Types 0, and Types 1 (for a more comprehensive listing please refer to our technical report), TreeVersity was designed to support tree comparison types 1 to 4.

Technical Requirements

TreeVersity was developed using modern web development programming tools and therefore requires a modern browser (IE9+, or recent versions of Chrome, Firefox and Safari). TreeVersity relies on the following technologies:

On the server side:

• Django for the application server, and therefore Python as a programming language on the server.
• PostgreSQL 9+ for the database
• Apache 2+ as the web server

On the client side:

• D3 for visualizations and their animations, and therefore Javascript as a programming language for the browser
• Bootstrap 2.0.2 for the look & feel
• JQuery 1.8.0 for some event handling and DOM manipulation
• JQuery UI 1.8.23 for some of the control widgets
• CrossFilter 1.0.3 for fast data queries on the browser, with some customizations to support NaN values, filters on multiple ranges and ascending rankings
• LESS for the style-sheets

Acknowledgements

TreeVersity was developed by John Alexis Guerra Gómez under the supervision of Dr. Ben Shneiderman and Dr. Catherine Plaisant, and under the design advise of Audra Buck-Coleman

TreeVersity was partially funded thanks to the support of the Fulbright Science and Technology Scholarship, the Center for Advanced Transportation Technology Laboratory and the Center for Integrated Transportation Systems Management (a Tier 1 Transportation Center at the University of Maryland)