News Details | HEAL

First Place in Symbolic Regression Competition

Apr 18, 2010

We are proud to announce that team HEAL achieved the first place at the symbolic regression competition of the EvoCompetitions side event of Evo* 2010. The goal of the competition was to create a prediction-model for a variable that is expensive to measure based on 57 other variables that can be easily measured based on data from a real world chemical process.

In the modeling process we tried to find a simple model that can be readily interpreted and validated by the domain experts. We put more emphasis on the interpretability, the prediction error of our final model is thus relatively large.

To generate the models for this competition we used an internal pre-release candidate of our software framework HeuristicLab that will be released real soon now. The fact that we achieved first place in this competiton and also second place in the financial time series competition shows that HeuristicLab is powerful and flexible enough to achieve competitive results for real world applications with minimal preparation effort.

Finally we would like to thank the organizers of the symbolic regression competition and of the financial time series prediction competition for making the datasets available for reseach purposes and the EvoCompetitions chair Anna I Esparcia-Alcázar for the organization of the event.