Bagged Polynomial Regression and Neural Networks
Researchers have developed Bagged Polynomial Regression with Random Projections (BPR), an ensemble method that averages regularized polynomial models. This approach aims to provide an auditable alternative to neural networks for high-dimensional prediction tasks, particularly in climate and environmental applications. BPR matches neural network accuracy in crop classification while offering greater transparency and interpretability through diagnostic tools. AI
IMPACT Offers a more interpretable alternative for complex prediction tasks, potentially aiding adoption in regulated or sensitive domains.