Enhancing Data Visualization Workflow through Visualization Recommendations

Abstract

As the volume and complexity of data continues to grow rapidly across various fields, the demand for effective data visualizations is even more important. Visualization plays a key role in data analysis, enabling users to better understand complex datasets and derive meaningful insights. However, selecting suitable visual representations remains a non-trivial task – particularly in notebook-based environments where users explore data in open-ended and iterative ways. While their primary focus lies on analytical reasoning, the additional cognitive effort required for data transformation or visual mapping decisions can disrupt their workflow and slow down the insight generation process. This thesis proposes an approach to enhance data visualization workflows by providing a chart-driven approach. Developed through a user-centered design process, the solution addresses different user needs and levels of expertise. A preliminary user study was conducted to understand users work contexts, challenges and expectations. Based on these insights, as well as a comprehensive literature and market analysis, an interactive low-fidelity prototype concept was designed to facilitate a more intuitive and supportive data visualization workflow. The prototype was integrated into a domain-specific notebook-based environment in the field of observability software and was evaluated through task-based testing and qualitative user feedback. The results show that participants with different levels of expertise perceived the concept as valuable, allowing them to concentrate on their analysis rather than on technical details. This work contributes to the design of mixed-initiative visualization recommendation systems.

Date of Award	2025
Original language	English
Supervisor	Mandy Keck (Supervisor)