Complexity, interpretability and robustness of GP-based feature engineering in remote sensing

João E. Batista, Adam Kotaro Pindur, Ana I. R. Cabral, Hitoshi Iba, Sara Silva

Abstract: Highlights•This work shows that different classifiers have different feature engineering needs.•We discuss the correlation between model complexity and performance.•We propose a new functional complexity metric for classification datasets.•We provide models to detect cocoa agroforest and to forecast forest degradation.•The dataset used to forecast forest degradation is publicly released on GitHub.