Visual Analytics

Geovisual Analytics for Social Media Data

3502.1 - Paper, Pen and Funny Doodles – How to Integrate Network Mining into Geovisualization Tools

Monday, July 3
1:30 PM - 1:50 PM
Location: Delaware B

This paper investigates the feasibility of integrating a sophisticated network data mining technique into SensePlace3 (SP3), a web-based geovisual analytics environment. The core contribution of this paper is a user study that determines whether an analyst with minimal background can comprehend the network data modeling metaphors employed by the resulting system, whether they can employ said metaphors to explore spatial data, and whether they can interpret the results of such spatial analysis correctly. This study confirms that all of the above is, indeed, possible, and provide empirical evidence about the importance of a hands-on tutorial and a graphical approach to explaining data modeling metaphors in the successful adoption of advanced data mining techniques.

This study focuses on an extension to SP3, a large, multi-year effort that aims to combine interactive visual interfaces with computational modeling techniques to help the analyst achieve insight about geospatial social media data. More specifically, we assess the usability and utility of Heterogeneous Information Network Mining (HINM) [1] — a computational technique that recently came into prominence. SP3 works with Twitter data that contains a large number of linked entities that lend themselves naturally to network-based representation and analysis. The extensive body of literature documenting the success of HINM across a number of application areas [2,3,4] makes HINM even more appealing as a candidate for inclusion in the SP3 toolkit.

Despite all of the promise of HINM, the complexity of the data modeling metaphors employed by this approach represents a significant adoption challenge. The notion of the heterogeneous network itself, the concept of “meta-paths”, the complexity of the algorithms involved in data analysis as well as the lack of obvious visual metaphors for use in (geo)visual workflow has thus far restricted this technique to the data-mining community with significant technical background.

The main contribution of this paper is a clear demonstration (based on a user study with 11 participants) that HINM metaphors can be used in a web-based geovisual analytics environment with minimal (under 15 minutes) amount of training for the analyst. Besides completing a set of analytical tasks based on real-world data, study participants left overwhelmingly positive feedback in terms of both Likert scale and open-ended questionnaire responses.

A second contribution is a realistic case study that demonstrates that HINM produces a clear and a positive difference in the outcome of spatial analysis. Although this study did not aim to validate HINM as a data modeling approach (existing literature contains considerable evidence for this), the results of the case study demonstrate that HINM is a viable geospatial analysis tool.

A third contribution is the design of a protocol for a geovisual analytics user study that was based on and improves upon the current state of the art. This protocol includes a walk-through, a hands-on tutorial, as well as a set of realistic data analysis tasks. Such protocols are rare — most studies reviewed in this paper fail to provide sufficient details for study replication or comparison work, making the proposed protocol an important contribution to the research on methodology of such studies.

1. Sun Y, Han J (2012) "Mining heterogeneous information networks: principles and methodologies." Synth Lect Data Min Knowl Discov 3:1–159.
2. Sun, Y, Han J (2013) "Meta-Path-Based Search and Mining in Heterogeneous Information Networks." Tsinghua Science and Technology 18 (4).
3. Shen W, Han J, Wang J (2014) "A probabilistic model for linking named entities in web text with heterogeneous information networks." ACM Press, 1199–1210.
4. Sun Y, Barber R, Gupta M, Aggarwal CC, Han J (2011) "Co-author Relationship Prediction in Heterogeneous Bibliographic Networks." Advances in Social Networks Analysis and Mining (ASONAM) 121–128.

Alexander Savelyev

Assistant Professor
Texas State University

Alexander (Sasha) Savelyev is a cartography and geovisualization scholar who focuses on issues of text (geo)visualization, Big Data visualization, user study design and human cognition in visual environments.


Send Email for Alexander Savelyev

Alan M. MacEachren

Professor of Geography
Penn State University

Alan M. MacEachren is Professor of Geography, Affiliate Professor of Information Sciences & Technology, and Director of the GeoVISTA Center at the Pennsylvania State University. MacEachren’s research foci include: geovisual analytics, cartography, geovisualization, geographical information retrieval/geoparsing, visual semiotics, spatial cognition, human-centered systems, user-centered design, geocollaboration. He is author of How Maps Work: Representation, Visualization and Design and Some Truth with Maps, as well as co-editor of additional books and journal special issues. He chaired the ICA Commission on Visualization (1999-2005). MacEachren is an Honorary Fellow of the ICA (2005) and was elected as a Fellow of AAAS in 2014.


Send Email for Alan MacEachren

Daria Hollenstein

Research Associate
FHNW University of Applied Sciences and Arts Northwestern Switzerland

Daria Hollenstein is a Research Associate at the Institute of Geomatics Engineering at the FHNW University of Applied Sciences and Arts Northwestern Switzerland in Muttenz (Basel). Currently, she works for the Geo-visualization and Visual Analytics group at the Institute. She has graduated from the University of Basel with a Master’s degree in Archaeology.


Send Email for Daria Hollenstein


3502.1 - Paper, Pen and Funny Doodles – How to Integrate Network Mining into Geovisualization Tools

Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for Paper, Pen and Funny Doodles – How to Integrate Network Mining into Geovisualization Tools