Project: Improving maintainability of data transformation graphs using visualization
Description
Companies often integrate their data by extracting information from different sources and applying various
transformations to receive their desired output (also referred to as ETL). Maintenance on the transformation graphs applying these transformations is performed by a small number of domain experts that often
have insufficient tools to do so. We introduce a novel approach that improves the maintainability of transformation graphs by enabling domain experts to recreate their own mental models using a graph hierarchy.
By presenting this hierarchy to others through a mental-map-preserving layered graph layout, users can
collaborate on a unified understanding of their transformation graph, enabling them to iteratively improve
its design and quickly correct errors. Using two real-world use cases and a qualitative analysis, we found
that collaboration on a unified mental model is very effective at improving maintainability. Additionally,
although finding errors has an initial learning curve for some users, it saves all of them a lot of time.
Details
- Student
-
RARuud Andriessen
- Supervisor
-
Jack van Wijk
- External location
- ProcessGold - Eindhoven
- Link
- Thesis