COMET-Project SmartDD (2019-2022)
Smart Data Discovery
Due to the constantly increasing amount of data sources, manual data analysis is reaching its limits in many areas. The implementation of semi-automatic data analyses is seen as a solution in the area of IoT and cyber physical production systems (CPPS). Current tools, however, mainly support approaches and methods that come from the business intelligence area. The aim of this project is to develop smart data discovery approaches specifically for IoT and CPPS. For this purpose, dedicated methods of machine learning will be developed that allow the inference of mechanistic process understanding in order to provide the best possible support for process analysis and process optimization in its industrial environment.
Whether it pertains to learning causes (causal discovery) of a disease, or effects of specific treatments in medical experiments, or finding the root cause in complex industrial processes, all tend to involve analysis of available data in order to understand why and how things happen.
In addition to typical operational data (e.g., sensor data readings), systems also record traces of execution in log files. This kind of event data is not commonly used, although it might contain valuable information.
- Efficient structural learning methods and their parallelized implementations (GPU/MultiCore)
- Robust causal discovery methods for industrial applications verified for application use cases provided by company partners
- Methods for assessing the reliability of predictions and their prototypical implementations
- Methods for model interpretation, model output explanation and integration of domain knowledge in learning using expert knowledge and/or domain ontologies
- Methods and tools for integrating process mining and causal discovery for enhanced process understanding
This project is subsidized in the frame of COMET – Competence Centers for Excellent Technologies by BMK, BMDW, State of Upper Austria and its scientific partners. The COMET program is handled by FFG.