Generic Methods

The project puts emphasis on a cross-fertilization paradigm. The goal is to translate specific methods identified within certain use cases into generic methods and tools. These will then become available to other use cases within the project and later to a wider audience of users within the Helmholtz Association of German Research Centers and beyond.

For each of the use cases, a set of major Scientific Big Data Analytics (SBDA) techniques will be implemented in an optimized manner. The goal here is to achieve convergence of domain and infrastructure software. The exchange of successful techniques among the use cases will exploit synergies and provide motivation for standardization and generalization of such techniques within the Helmholtz Data Federation (HDF).

The close collaboration of the scientists who are involved in the integration and analysis of the data ensures that access to all necessary information on the data is available for each use case. In fact, the combination and analysis of complex data from a diversity of sources might itself provide new opportunities to generate novel insights, e.g. combining medical cohort data with terrestrial systems data.

The following methods are used and researched within the project:

• Unsupervised cluster analysis
• Supervised classification problems
• Forecast regression
• Sequence mining
• Data assimilation
• Model and parameter optimization

Results

The firsts results of these activities are available as two open sources libraries: HeAT and Hyppopy.