Our research is inspired by the challenges of changing environments. We develop methods for model learning, model adaption and change monitoring on streams. We work with timestamped data for applications in medicine and healthcare and for business. Next to analysing observational data, we also analyse experimental data. Furthermore, we conduct experiments ourselves on the interaction between human and machine (e.g. for active learning) to identify difficult tasks and to assess human uncertainty. The latter takes place in close cooperation with the Technical University Ilmenau and the Chemnitz University of Technology within the CHIM network.
In the thematic area Medical Mining, we develop mining methods for the analysis of cohort data from epidemiological and clinical studies, and of observational data from clinical databases and mHealth apps. A major challenge in this area is missingness, especially in the temporal mHealth data. We address this challenge by developing parsimonious methods that demand few instances, that take the cost of feature acquisition into account, that acquire features actively and that assess similarity among instances despite large differences in the time series lengths.
In the thematic area Experiments I, we develop temporal mining methods to acquire insights on the behaviour of participants. The emphasis is less on prediction and more on identifying the salient motifs that explain the actions of the participants and their transitions during the experiment.
In the thematic area Experiments II, we design experiments on human-machine interaction, where a ‘machine’ can be a physical entity (as in an industrial setting) or an active learning component (as for the annotation of medical records). We focus on distinguishing between easy and difficult tasks and on assessing human uncertainty.
A detailed list of current and completed projects can be found on the reseach portal of Saxony-Anhalt, click here (in German).