Applied machine learning in exploring key features of crayfish populations

Abstract

Uniform Manifold Approximation and Projection (UMAP) is a nonlinear dimension reduction technique based on manifold learning. It is specifically designed to achieve a balance between the global and local structure when embedding data points. We applied this method to morphometric features in populations of the noble crayfish, a freshwater species recognized as both a keystone species and an ecosystem engineer, as well as an indicator of good water quality, with unquestionable cultural and economic value to humans. Our results show that the CLL parameter most contributed to the differences and grouped the investigated specimens into seven clusters, along with ROL and ABL parameters. The parameters associated with the claws also exhibited a considerable influence on differentiation.