Berlin 2015 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

DY: Fachverband Dynamik und Statistische Physik

DY 60: Poster - complex systems and data analysis

DY 60.1: Poster

Thursday, March 19, 2015, 16:00–18:00, Poster A

Modelling and prediction using symbolic regression by genetic programming — •Markus Quade, Markus Abel, and Ruggero Vasile — Ambrosys GmbH, Potsdam, Deutschland

Machine learning assisted modelling is used successfully in a broad variety of systems. However, the models are often not physically interpretable. Symbolic regression, in contrast, yields formulas as results, which can be analyzed by means of dynamical systems methods.

Specifically, we use evolutionary principles to explore possible models to forecast the production of green energy by genetic programming. In order to find suitable models in acceptable time, we extend the standard genetic programming method.

One advantage of this approach is that it is in principle not biased by human perception. Of course, prior knowledge about the system investigated should be used by providing building blocks and a goodness of fit measure. Finding a model means searching for a best fit in function and/or parameter space.

We illustrate our ideas in detail with real measurements and the corresponding prediction.