New Methods for Activity-based Model Analysis and Visualization
B Stabler, J Freedman, P Vovsha, Parsons Brinckerhoff, US
This paper describes a model visualization system developed for the Atlanta, USA regional government that merges the latest ideas in travel forecasting, individual behavior research, and visual analytics.
This paper describes new methods for activity-based model (ABM) analysis and visualization. For the Atlanta, USA regional government, a model visualization dashboard and database backend was developed that merges the latest ideas in travel forecasting, individual behavior research, and visual analytics. The Atlanta Regional Commission's (ARC) new microsimulation ABM produces gigabytes of data that is analogous to a comprehensive household survey for the entire regional population. This wealth of data can be difficult to process, analyze and glean useful information from in a timely manner. ARC and its consultant staff have worked on visualization of model results during specific projects, but these efforts have been disjointed and partial. To make the model more accessible, and to improve the speed at which a useful and meaningful summary can be generated, a comprehensive end-user focused model visualization dashboard was developed. The proposed structure is very flexible and can accommodate many types and dimensions of analysis. However, over the recent years, some important and most frequently used forms of analysis and visualization have become clear.
The project was split into two distinct tasks.
The first task was to develop a reporting/visualization database using ARC's enterprise SQL Server database. This database centralized and structured the model outputs in a consistent manner that could be automated to follow a model run. Properly structuring the data in a generic/normalized fashion allowed for the development of a tool to query/mine the data. In addition, using a database system that features parallel query capabilities means a tremendous amount of data processing can be done very quickly, which is essential for exploratory data visualization with a database containing millions of person and household records.
The second task was to develop an Adobe Flash interactive dashboard application that connects to the visualization database, allows for standard and custom data queries, and presents results in dynamic tables, charts, animations, and maps. The dashboard includes standard interactive visuals such as interactive bar charts and maps, but also innovative visuals from the data visualization/visual analytics community such as animated time use charts for exploring person time use for every hour of the day by activity and person type, interactive tracing of tours and individual daily patterns in time and space, treemaps for the display and interactive investigation of hierarchical data (such as nested mode choice model results), radar charts for the comparison of performance measures across alternatives and/or geographic areas, and the ability to difference results between scenarios.
The practical value of the new tool is the speed at which model results can be investigated, patterns revealed, and scenario "stories" put together to share with others. The dynamic visualization dashboard has already proven its worth by being used for regional transportation planning model run analysis. With this tool, an end-user of an ABM can take a full advantage of the wide spectrum of possible dimensions for activity and travel analysis that are not possible with conventional 4-step models.
The paper and presentation will include various practical examples of analysis and visualization for different transportation projects and policies.
Association for European Transport