Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Visual exploration of large scatter plot matrices by pattern recommendation based on eye tracking

: Shao, Lin; Silva, Nelson; Eggeling, Eva; Schreck, Tobias


Glowacka, Dorota (Ed.) ; Association for Computing Machinery -ACM-; Association for Computing Machinery -ACM-, Special Interest Group on Artificial Intelligence -SIGART-; Association for Computing Machinery -ACM-, Special Interest Group on Computer and Human Interaction -SIGCHI-:
ESIDA 2017, ACM Workshop on Exploratory Search and Interactive Data Analytics. Proceedings : Limassol, Cyprus, March 13, 2017
New York: ACM, 2017
ISBN: 978-1-4503-4903-1
Workshop on Exploratory Search and Interactive Data Analytics <2017, Limassol>
International Conference on Intelligent User Interfaces (IUI) <22, 2017, Limassol>
Conference Paper
Fraunhofer Austria ()
user/machine system; human information processing; Information storage and retrieval; eye tracking; Guiding Theme: Visual Computing as a Service; Research Area: Human computer interaction (HCI)

The Scatter Plot Matrix (SPLOM) is a well-known technique for visual analysis of high-dimensional data. However, one problem of large SPLOMs is that typically not all views are potentially relevant to a given analysis task or user. The matrix itself may contain structured patterns across the dimensions, which could interfere with the investigation for unexplored views. We introduce a new concept and prototype implementation for an interactive recommender system supporting the exploration of large SPLOMs based on indirectly obtained user feedback from user eye tracking. Our system records the patterns that are currently under exploration based on gaze times, recommending areas of the SPLOM containing potentially new, unseen patterns for successive exploration. We use an image-based dissimilarity measure to recommend patterns that are visually dissimilar to previously seen ones, to guide the exploration in large SPLOMs. The dynamic exploration process is visualized by an analysis provenance heatmap, which captures the duration on explored and recommended SPLOM areas. We demonstrate our exploration process by a user experiment, showing the indirectly controlled recommender system achieves higher pattern recall as compared to fully interactive navigation using mouse operations.