“…Lately, machine learning methods find one's way into individual claims reserving, allowing for more flexible regression structures. Some recent papers are based on regression trees and gradient boosting, see Wüthrich (2018), Lopez et al (2019), Lopez & Milhaud (2021), De Felice & Moriconi (2019 and Duval & Pigeon (2019); others are based on non-parametric and kernel methods, see Rosenlund (2012), Bischofberger et al (2019), Baudry & Robert (2019), or on neural networks, see Gabrielli (2020), Kuo (2020) and Delong & Wüthrich (2020). Surprisingly, many of these approaches pay little attention to the data itself, but they describe the methods used in much detail.…”