Machine learning

We help our clients to make use of their data through state-of-the-art machine learning technology. We interpret the data, identify suitable predictive models, and quantify the model confidence.
Gaussian process regression intrinsically supports quantifying the model confidence.

With advances in IT, monitoring and sensor technology, an increasing amount of data on the parameters and performance of engineering systems are collected. Data is collected at different design and operation phases of engineering systems and processes. In some cases, a vast amount of data might be available (also known as big data), in other cases the data might be limited and/or even incomplete. It is crucial that data processing and analysis are performed effectively in a systematic manner. Machine learning algorithms provide the tools needed to extract information from and make optimal use of available data. They can identify and employ patterns in the recorded data, even if an underlying process-based model is unavailable.

We help our clients to integrate machine learning soundly into the probabilistic modeling of uncertain model parameters. We also support our clients in developing machine learning-based predictive models of the underlying physical processes. State-of-the-art machine learning techniques enable exploiting the full economic potential that the data offers for effective decision-making. We employ enhanced graphical data representations and visualizations to facilitate the communication of analysis results.

Observed data samples and joint PDF derived through data analysis.