B0989
Title: Explaining spatial regression random forest
Authors: Natalia Golini - University of Turin (Italy) [presenting]
Luca Patelli - University of Pavia (Italy)
Rosaria Ignaccolo - University of Turin (Italy)
Michela Cameletti - Universita degli Studi di Bergamo (Italy)
Abstract: Random forests (RF) is a widely used supervised machine learning algorithm in geospatial/point-referenced applications due to its flexible nature and strong predictive performance. However, RF is considered a black box model since it prevents grasping how predictors are combined to generate the response variable predictions. The lack of interpretability becomes especially problematic when decision-making requires understanding the relationship between the response and predictors. The aim is to explain regression RF specifically designed for spatially dependent data by extracting a stable, short, and easy-to-understand set of rules. A spatial extension of SIRUS is proposed, a regression rule algorithm designed to extract rules from classical random forests. The approach is illustrated through the analysis of a spatial dataset.
