In the research area Data Science, we focus on the development and application of data-driven methods for solving biological, medical, and technical problems. In close cooperation with our partners from research and industry, we develop algorithms for preprocessing and analyzing data using statistical methods and machine learning, as well as visualizing and interpreting the results. For example, within the project LeiVMed Online we develop a platform for benchmarking and visualizing data of treatments of patients in hospitals and developing prediction models for the outcome of treatments.
In many of our projects, we see the need for customized data processing pipelines due to the heterogeneity and complexity of data structures in real world processes and systems. We use applied statistics as well as numerous machine learning approaches, including black box methods (deep neural nets, random forests, gradient boosted trees, etc.) and white box methods (symbolic regression by genetic programming). We build our knowledge discovery pipelines on a variety of different frameworks, especially python scikit-learn, tensorflow, pytorch, MATLAB, and HeuristicLab.
|2019||Renate Haselgrübler, Peter Lanzerstorfer, Clemens Röhrl, Flora Stübl, Jonas Schurr, Bettina Schwarzinger, Clemens Schwarzinger, Mario Brameshuber, Stefan Wieser, Stephan M. Winkler, and Julian Weghuber:||Hypolipidemic effects of herbal extracts by reduction of adipocyte differentiation, intracellular neutral lipid content, lipolysis, fatty acid exchange and lipid droplet motility||Scientific Reports||View Article|
|2018||Stephan M. Winkler:||Evolutionary Computation and Symbolic Regression in Scientific Modeling||, Johannes Kepler University Linz, Computer Science, Habilitation Thesis|
The österreichische Bergkräutergenossenschaft obtains its raw materials (mainly herbs) from various farmers mainly from the Mühlviertel region. However, microbial contamination often occurs with these materials, which leads to high further processing costs. In the project MicoRed we try to reduce this microbial contamination with the help of applied statistics and machine learning. Here different machine learning methods are applied. (e.g. Random Forest, Gradient Boosting Trees, Genetic Programming, Neural networks). For this, a web application was implemented in which farmers collect data per batch about the harvest, the field, the drying, the weather. This data is then used for the above-mentioned evaluations in the application and presented to the farmers. For example, a farmer receives the information whether there is a possible germination in already harvested herbs and which part of the data is responsible for this.
Prof.(FH) DI Dr. Stephan Winkler
Stefan Anlauf Bsc.
Andreas Haghofer Msc.
11/2017 – Present
FFoQSI (Austrian Competence Centre for Feed and Food Quality Safety and Innovation) University of Applied Sciences, Upper Austria, Hagenberg Campus
Software technology and application; Data Science and Data Engineering