Development of a software framework for constructing species distribution models to assess the limits of the ecological valence of introduced species in relation to the main climatic factors

The results of the stages of development and testing of the species distribution modeling scheme (SDM) using the Biomod 2 and Targets packages in the R programming language are presented. Three species of the genus Aloe L. were selected as objects of research from the collections of the Donetsk Botanical Garden growing in South Africa, for the analysis of which the Global Biodiversity Information Facility (GBIF) database was used. The simulation included the use of an ensemble of various machine learning algorithms, such as the generalized boosting model (GBM), multiple adaptive regression splines (MARS) and artificial neural network (ANN). This allowed us to achieve high quality indicators of the models, with KAPPA in the range from 0.657 to 0.917 and TSS from 0.74 to 0.943. The developed model has demonstrated high accuracy and reliability, which makes it suitable for further scaling and use in large-scale analyses.