Enhancing empirical modelling in environmental science with knowledge discovery and genetic programming

Abstract

: Genetic programming (GP) has shown great promise in empirical modelling for environmental science, particularly in complex systems such as climate, flood, and environmental modelling. However, the success of GP largely depends on the quality and quantity of data used for training. In this regard, knowledge discovery (KD) can significantly improve GP’s ability to model complex interactions (Grin and Gandomi 2021). KD is the process of discovering new knowledge or insights from existing data, often through data mining and machine learning techniques. KD can be used in conjunction with GP to identify relevant variables, patterns, and interactions within a dataset, which can then be used to improve the accuracy and generalization of GP models. By discovering new knowledge, KD can also help GP to avoid overfitting and capture more complex relationships between variables.