SOMevidence - Evidence Workflows for Self-Organizing Map Classification

Provides explicit data-design objects, leakage-safe preprocessing, design-aware resampling, reproducible ensembles of self-organizing maps, and evidence-oriented diagnostics for ecological, environmental and evolutionary applications. The package separates continuous-representation diagnostics from the defensibility of a hard partition and supports explicit abstention when evidence is insufficient. Self-organizing maps follow Kohonen (1982) <doi:10.1007/BF00337288>, using the 'kohonen' implementation described by Wehrens and Kruisselbrink (2018) <doi:10.18637/jss.v087.i07>.