Title: Augmenting Exploratory Data Analysis with Visualization Recommendation

Advisor: Jeff Heer

Supervisor Committee: Jeff Heer (Chair), Jevin West (GSR, iSchool), Bill Howe, and Jock Mackinlay (Tableau)

Abstract: 

Exploratory data analysis is one of the key activities for understanding
and discovering new insights from the data. The exploratory data analysis
process involves both open-ended exploration and focused question answering.
Existing visualization tools support a variety of charts for question answering.
However, they typically require manual chart specification, making them tedious
for open-ended exploration where systematic data coverage should be achieved.
Without discipline and time, analysts may overlook important insights in the data, such as potentially confounding factors and data quality issues, or
prematurely fixate on specific questions or hypothesis.

To help analyst perform rapid and systematic data exploration,
this dissertation contributes the design of visualization tools
that complement manual specification with recommendation.

In this dissertation, we first present an interview study with data analysts from both professional and academic settings.
We characterize the process of exploratory data analysis and
key challenges in the process, as well as discuss design implications
for data exploration tools.

We then describe new formal languages and systems that are foundations for
chart specification and recommendation. The Vega-Lite visualization
grammar provides a representation for specifying and reasoning about charts.
The CompassQL query language and recommender engine provide
a generalizable framework for chart recommendation via queries
over the space of visualizations.

With these formal languages, we then used the iterative design process to
develop and study new recommendation-powered data visualization interfaces for
exploratory analysis. Voyager enables data exploration via
browsing of recommended charts. Our user study, which compared Voyager
with a traditional chart authoring tool, indicated the complementary benefits of
manual authoring and recommendation browsing.
Inspired by the Voyager study result, Voyager 2 blends manual and
automated chart authoring in a single tool to facilitate rapid and systematic
data exploration while preserving users' flexibility to author a broad range of charts.

All of these systems have been released as open-source projects and adopted
in both research and professional data science communities.

Place: 
CSE 503
When: 
Tuesday, July 10, 2018 - 10:00 to Thursday, March 28, 2024 - 06:58