Interactive Visual Analytics as a Service

 

We address the following research questions: How can we develop generic multidimensional Visual Analytics services to support specific domain analysis? How can we empower subject-matter experts, decision makers, and the general public (assumed non-programmers) with usable interactive visual analytics? How do we provide ad hoc visual analytics to explore and discover new and hidden patterns in Big Data sets? How do we support live exploratory visual analysis through interactive gesture-based and multi-touch filtering, zooming, and cross referencing? The hypothesis is that we can develop highly interactive visual analytics methods and tools that make Big Data analytics accessible to non-programmer domain experts as a service.

Supporting Exploration and Ad Hoc Analysis

Research in data analytics is often model-driven, where the given data model is taken as the starting point for analysis. Also many visualization methods take the data model as the basis for analysis, e.g. let's meta-godals of interpretation and trust drive and inform the design of model-driven visualizations. Intuitive interation for visual analytics is needed for non-programmers to analyze Big Data. We need to support exploration through live zoom and dynamic filtering based on multi-touch brushing to fully interactive interfaces with collaborative gestural zoom, filtering and overiews, e.g. for decision makers in a boardroom. We will develop methods supporting humans to explore and perform ad-hoc analysis of data patterns, which can be used as input for algorithmic models or ML, and be targeted to handle problems identified in the domain cases.

 

Visual Analytics for Multidimensional data

Much data today is produced by mobile and stationary sensing systems or rich self-service systems with possibilities for tagging data along multiple dimensions. This calls for general online visual analysis services that provide 1) handling of temporal, spatial or thematic dimensions in data, and 2) global pictures of sensor data for human activity or location tracking, 3) trend analysis in regards to local and global aspects. Such tools build on research in statistical aggregation, data mining, pattern extraction and spatio-temporal data structures and query processing. Nevertheless, it will also need to utilize and deal with semantic annotation (time, location, type of source, etc.) of data, originating from diverse sensing sources, exhibiting noise, irregular sampling and varying accuracies. Tools need to offer easy entry of different multi-dimensional data sets and in turn easily interpretable global views of multi-dimensional data, while also highlighting local phenomena.

Description

With the huge growth in sensor and human registered data, we need visualization tools as plug-and-play service components leveraging from custom applications to reusable tools applicable across different domains. We will look into:

Supporting Exploration and Ad Hoc Analysis
Visual Analytics for Multidimensional data