Big Data and Service Science


We dedicate our work to research on decision-making processes based on facts and data regarding the planning, controlling, and designing of complex enterprise (IT) services and systems.

In enterprises, massive amounts of historical sensor, process monitoring, and customer usage data is collected. Theoretically, this data can be used for prediction and optimisation as well as for automated or partly automated controlling.

Unfortunately, the complexity resulting from the analysis of large amounts of multivariate, fine-grained data, leads to the fact that dependencies and relationships within the data are not found, algorithms do not scale, and traditional static procedures as well as data mining techniques fail to work because of the well-known “curse of data dimensionality”.

Today, these problems are often referred to as big data challenges.

The result is that, in practice, data is mostly aggregated in a problem-oriented fashion, and for most problems the methods applied are rather simple and conservative heuristics or even manual approaches.

Big Data and Service Science at the FZI

At the FZI we approach these challenges from various perspectives. Firstly, we develop mathematical solutions to project high-dimensional data spaces onto a few key features that depict the basic structures with regard to their problem-specific developments and relations within the data. Based on the concise data at hand, we draft scalable decision models. Secondly, we work on efficient means of processing huge data streams with the aim of anticipating anomalies, structural breaks, or other kinds of patterns that require interventions. Thirdly, we develop models to calculate the economic impact of decisions and add coordination and market mechanisms to handle resulting complex economic issues.

Current R&D projects

In current R&D projects we focus on several topics. For instance, we make robust long-term prediction models concerning the demand and unit costs in large enterprises. Furthermore, we explore the detection of behavioural patterns in comprehensive corporate financial planning data leading to inefficiencies in the planning process. Two other research focuses are the high-dimensional segmentation of customers for campaigns in the telecommunications industry as well as the detection of complex patterns in smart meter monitoring data that reveal inefficiencies or failures. Additionally, we research on the design of dynamic service value networks by orchestrating adequate corporate or cloud services according to multi-facetted functional and non-functional criteria as well as enterprise-specific constraints.

Contact person