Learning Ensembles of Process-Based Models by Bagging of Random Library Samples

摘要: We propose a new method for learning ensembles of process-based models predictive modeling dynamic systems from data and knowledge. Previous work has shown that based on sampling (i.e., bagging), significantly improve performance models. However, this improvement comes at the cost substantial computational overhead needed learning. On other hand, methods constructing knowledge random library samples, RLS) allow efficient models, while maintaining their long-term performance. This paper aims checking conjecture whether combination these potential further improvements. The proposed method, bagging samples combines afore-mentioned approaches in terms both apply to evaluate its set automated tasks two lake ecosystems population dynamics. experimental results serve identify optimal design decisions regarding as well asses ability. show such outperform single model, but also each (bagging) (RLS).