Jimmy Johansson

Revealing structure within clustered parallel coordinates displays

In order to gain insight into multivariate data, complex structures must be analysed and understood. Parallel coordinates is an excellent tool for visualizing this type of data but has its limitations. This paper deals with one of its main limitations - how to visualize a large number of data items without hiding the inherent structure they constitute. We solve this problem by constructing clusters and using high precision textures to represent them. We also use transfer functions that operate on the high precision textures in order to highlight different aspects of the cluster characteristics. Providing predefined transfer functions as well as the support to draw customized transfer functions makes it possible to extract different aspects of the data. We also show how feature animation can be used as guidance when simultaneously analysing several clusters. This technique makes it possible to visually represent statistical information about clusters and thus guides the user, making the analysis process more efficient.

An heuristic set for evaluation in information visualization

Evaluation is a key research challenge within the international Information Visualization (InfoVis) community, and Heuristic Evaluation is one recognized method. Various sets of heuristics have been proposed but there remains no consensus as to which heuristics are most useful for addressing aspects specific to the complex interactive visual displays used in modern InfoVis systems. This paper presents a first effort to empirically determine a new set of such general heuristics tailored for Heuristic Evaluation of common and important usability problems in InfoVis techniques. Participants in the study rated how well a total of 63 heuristics from 6 earlier published heuristic sets could explain a collection of 74 usability problems derived from earlier InfoVis evaluations. The results were used to synthesize 10 heuristics that, as a set, provided the highest explanatory coverage. The paper also stresses the challenges for future research to validate and further improve upon this set.

ActiviTree: Interactive Visual Exploration of Sequences in Event-Based Data Using Graph Similarity

The identification of significant sequences in large and complex event-based temporal data is a challenging problem with applications in many areas of todays information intensive society. Pure visual representations can be used for the analysis, but are constrained to small data sets. Algorithmic search mechanisms used for larger data sets become expensive as the data size increases and typically focus on frequency of occurrence to reduce the computational complexity, often overlooking important infrequent sequences and outliers. In this paper we introduce an interactive visual data mining approach based on an adaptation of techniques developed for Web searching, combined with an intuitive visual interface, to facilitate user-centred exploration of the data and identification of sequences significant to that user. The search algorithm used in the exploration executes in negligible time, even for large data, and so no pre-processing of the selected data is required, making this a completely interactive experience for the user. Our particular application area is social science diary data but the technique is applicable across many other disciplines.

Revealing structure in visualizations of dense 2D and 3D parallel coordinates

Parallel coordinates is a well-known technique used for visualization of multivariate data. When the size of the data sets increases the parallel coordinates display results in an image far too cluttered to perceive any structure. We tackle this problem by constructing high-precision textures to represent the data. By using transfer functions that operate on the high-precision textures, it is possible to highlight different aspects of the entire data set or clusters of the data. Our methods are implemented in both standard 2D parallel coordinates and 3D multi-relational parallel coordinates. Furthermore, when visualizing a larger number of clusters, a technique called ‘feature animation’ may be used as guidance by presenting various cluster statistics. A case study is also performed to illustrate the analysis process when analysing large multivariate data sets using our proposed techniques.

The GAV Toolkit for Multiple Linked Views

Implementing InfoVis multivariate data tools, timelinked coordinated views and visual dynamic queries with conditioning from scratch is not a simple programming task. Our research objective is to develop a generic GeoAnalytics visualization (GAV) component toolkit, based on the principles behind visual analytics (VA), for dynamically exploring time-varying, geographically referenced and multivariate attributes simultaneously. GAV includes components based on a synergy of technologies from information visualization, geovisualization and scientific visualization. Our research concentrates on improving visual user interfaces (VUI) methods and trying to extend existing visual representation techniques. The effectiveness of our proposed component toolkit and framework is demonstrated in two customized applications GeoWizard analysing multivariate energy usage data for Swedish municipalities and MD-Explorer exploring multivariate data using novel interactive ternary diagrams. We use parallel coordinates with embedded visual inquiry methods that serves as a visual control panel for dynamically linked and coordinated views. Finally, discoveries made during the visual exploration process can be captured and organized in a format for later recall and communication to others.

Interaction Support for Visual Comparison Inspired by Natural Behavior

Visual comparison is an intrinsic part of interactive data exploration and analysis. The literature provides a large body of existing solutions that help users accomplish comparison tasks. These solutions are mostly of visual nature and custom-made for specific data. We ask the question if a more general support is possible by focusing on the interaction aspect of comparison tasks. As an answer to this question, we propose a novel interaction concept that is inspired by real-world behavior of people comparing information printed on paper. In line with real-world interaction, our approach supports users (1) in interactively specifying pieces of graphical information to be compared, (2) in flexibly arranging these pieces on the screen, and (3) in performing the actual comparison of side-by-side and overlapping arrangements of the graphical information. Complementary visual cues and add-ons further assist users in carrying out comparison tasks. Our concept and the integrated interaction techniques are generally applicable and can be coupled with different visualization techniques. We implemented an interactive prototype and conducted a qualitative user study to assess the concepts usefulness in the context of three different visualization techniques. The obtained feedback indicates that our interaction techniques mimic the natural behavior quite well, can be learned quickly, and are easy to apply to visual comparison tasks.

Task-based evaluation of multirelational 3D and standard 2D parallel coordinates

Multivariate data sets exist in a wide variety of fields and parallel coordinates visualizations are commonly used for analysing such data. This paper presents a usability evaluation where we compare three types of parallel coordinates visualization for exploratory analysis of multivariate data. We use a standard parallel coordinates display with manual permutation of axes, a standard parallel coordinates display with automatic permutation of axes, and a multi-relational 3D parallel coordinates display with manual permutation of axes. We investigate whether a 3D layout showing more relations simultaneously, but distorted by perspective effects, is advantageous when compared with a standard 2D layout. The evaluation is accomplished by means of an experiment comparing performance differences for a class of task known to be well-supported by parallel coordinates. Two levels of difficulty of the task are used and both require the user to find relationships between variables in a multivariate data set. Our results show that for the manual exploration of a complex interrelated multivariate data set, the user performance with multi-relational 3D parallel coordinates is significantly faster. In simpler tasks, however, the difference is negligible. The study adds to the body of work examining the utility of 3D representations and what properties of structure in 3D space can be successfully used in 3D representations of multivariate data.

Depth cues and density in temporal parallel coordinates

This paper introduces Temporal Density Parallel Coordinates (TDPC) and Depth Cue Parallel Coordinates (DCPC) which extend the standard 2D parallel coordinates technique to capture time-varying dynamics. The proposed techniques can be used to analyse temporal positions of data items as well as temporal positions of changes occurring using 2D displays. To represent temporal changes, polygons (instead of traditional lines) are rendered in parallel coordinates. The results presented show that rendering polygons is superior at revealing large temporal changes. Both TDPC and DCPC have been efficiently implemented on the GPU allowing the visualization of thousands of data items over thousands of time steps at interactive frame rates.

A screen space quality method for data abstraction

The rendering of large data sets can result in cluttered displays and non‐interactive update rates, leading to time consuming analyses. A straightforward solution is to reduce the number of items, thereby producing an abstraction of the data set. For the visual analysis to remain accurate, the graphical representation of the abstraction must preserve the significant features present in the original data. This paper presents a screen space quality method, based on distance transforms, that measures the visual quality of a data abstraction. This screen space measure is shown to better capture significant visual structures in data, compared with data space measures. The presented method is implemented on the GPU, allowing interactive creation of high quality graphical representations of multivariate data sets containing tens of thousands of items.

Integration of unsupervised clustering, interaction and parallel coordinates for the exploration of large multivariate data

Parallel coordinates are widely used in many applications for visualization of multivariate data. Because of the nature of parallel coordinates, the visualization technique is often used for data overview. However, when the number of tuples to be visualized becomes very large, this technique makes it difficult to distinguish the overall structure. In This work we present a novel technique which uses a classification approach, the self-organizing map (an unsupervised learning algorithm), to solve this problem by creating an initial clustering of the data. By initially only visualizing the resulting representational clusters, the inherited global structure can be shown. Using linked views and allowing the user to perform drill-down and filtering on these representations reveals the single data items without loss of context.