Datenanalyse und Visualisierung Prof. Dr. Daniel A. Keim

Scalable Visual Analytics of Movement

In vielen Anwendungsbereichen werden Daten mit Raum- und Zeitbezug in schnell wachsender Anzahl erhoben. Raum-Zeit-Daten und insbesondere Bewegungsdaten, beziehen geographischen Raum, Zeit und multidimensionale Attribute mit ein und stellen dadurch erhebliche Herausforderungen an die Analyse.

Das Ziel des Projekts ist, angemessene Analysemethoden zu entwickeln, welche visuelle, interaktive und algorithmische Techniken zu skalierbarer Analyse kombinieren. Die algorithmischen Techniken werden Trajektorien, Verkehrssituationen, Ereignisse und korrespondierende multidimensionale Attribute clustern, aggregieren und zusammenfassen.

Dadurch wird ein Überblick auf die Daten geschaffen und bedeutsame Muster von potentiellem Interesse extrahiert. Visuelle und interaktive Techniken erlauben dem Benutzer, automatisierte Algorithmen zu steuern und ihr Wissen in den Analyseprozess einzubringen. Zielanwendungen sind im Bereich GPS/Traffic Control, Delivery Tracing, CreditCard Fraud Detection, Cellphone Usage Analysis.
(ViAMoD: Visual Spatiotemporal Pattern Analysis of Movement and Event Data, Funded by DFG SPP 1335)

Density equalizing distortion

Visualizing large geo-demographical datasets using pixel-based techniques involves mapping the geo-spatial dimensions of a data-point to screen coordinates and appropriately encoding its statistical value by color. Analysis of such data is a great challenge. General tasks involve clustering, categorization and searching for patterns of interest for sociological or economic research. Available visual encodings and screen space limitations lead to over-plotting and hiding of patterns and clusters in densely populated areas, while sparsely populated areas waste space and draw the attention away from areas of interest. In the current paper, two new approaches (RadialScale and AngularScale) are introduced to create density-equalized maps, while preserving recognizable features and neighborhoods in the visualization. These approaches build the core of a multi-scaling technique based on local features of the data described as local minima and maxima of point density. Consequently, scaling is conducted several times around these features leading to more homogeneous distortions. Results are discussed on several real-world datasets. Evaluation shows that the proposed techniques outperform traditional ones regarding the homogeneity of resulting data distributions building a more appropriate bases for analytic purposes.

Fig.1 Distortion results for Europe dataset representing Wiki-Points of 5 selected languages. The languages shown are French, English, German, Portuguese and Spanish (from left to right on the color-scale, which is a 5 level qualitative color map from ColorBrewer). The map aims to show the dominance of one (or more) of these languages and diversification at certain areas.

Fig. 2 Distortion results for the USA dataset showing the average household income for 1999. The income is mapped to color (using a 11 class diverging color scale from ColorBrewer) reaching from red (low income) to blue (high income). The distortion shows constellations of cities and country side areas in a comparable size and highlight the cities' heterogeneous nature. New York city / Manhattan is enlarged (lower right corner) to show a better view on its diversity.

The process of distorting, as a transmission from an original map into a distorted map, is shown in the following short movie using the Europe dataset.

Generalized Scatter Plots

Scatter Plots are one of the most powerful and most widely used techniques for visual data exploration. A well-known problem is that scatter plots often have a high degree of overlap, which may occlude a significant portion of the data values shown. In this paper, we propose generalized scatter plots, which allow the visualization of large amounts of data to fit entirely into the display window without overlap. We discuss two variants: binned scatter plots and distorted scatter plots. The basic idea is to allow the analyst to optimize the degree of overlap, distortion, and binning to generate the best possible view. To allow an effective usage, we provide the capability to zoom smoothly between the traditional and our generalized scatter plots. We identify an optimization function which takes overlap and distortion of the visualization into account. We evaluate the generalized scatter plots according to this optimization function, and show that there usually exists a optimal compromise between overlap and distortion. Our generalized scatter plots have been applied successfully to a number of real-world IT services applications, such as servers performance monitoring, telephone service usage analysis, and financial data, demonstrating the benefits of the generalized scatter plots over traditional ones.

Generalized Telephone Service Scatter Plot without Overlap. X-axis shows duration in seconds, Y-axis shows the charges in dollar, the color represents the number of participants. Original data represented in upper left corner, overlap is stepwise reduced (left-right) and distortion is stepwise increased (up-down).

The transition from an original scatter plot to a overlap-optimized scatter plot (Generalized Scatter Plot) can be viewed here...

Growth Ring Maps: Spatiotemporal Analysis of Sensor Logs

Spatiotemporal analysis of sensor logs is a challenging research field due to three facts: a) traditional two-dimensional maps do not support multiple events to occur at the same spatial location, b) three-dimensional solutions introduce ambiguity and are hard to navigate, and c) map distortions to solve the overlap problem are unfamiliar to most users. This paper introduces a novel approach to represent spatial data changing over time by plotting a number of non-overlapping pixels, close to the sensor positions in a map. Thereby, we encode the amount of time that a subject spent at a particular sensor to the number of plotted pixels. Color is used in a twofold manner; while distinct colors distinguish between sensor nodes in different regions, the colors’ intensity is used as an indicator to the temporal property of the subjects’ activity. The resulting visualization technique, called Growth Ring Maps, enables users to find similarities and extract patterns of interest in spatiotemporal data by using humans’ perceptual abilities. We demonstrate the newly introduced technique on a dataset that shows the behavior of healthy and Alzheimer transgenic, male and female mice. We motivate the new technique by showing that the temporal analysis based on hierarchical clustering and the spatial analysis based on transition matrices only reveal limited results. Results and findings are cross-validated using multidimensional scaling. While the focus of this paper is to apply our visualization for monitoring animal behavior, the technique is also applicable for analyzing data, such as packet tracing, geographic monitoring of sales development, or mobile phone capacity planning.