The article presents a study aimed at developing methods and algorithms for analyzing spatial data for diagnosing the state of geosystems. It is shown that the combination of machine learning models into an ensemble makes it possible to increase the stability of the analyzing system: the accuracy of decisions made by the ensemble tends to tend to the accuracy of the most efficient monoclassifier of the system. The calculation and consolidation of territorial descriptors at the same time make it possible to reduce the dimensionality of the analyzed data, facilitate the allowable capacity of the machine learning model, increase its resistance to overfitting, and prevent a significant decrease in the classification accuracy within the framework of a specific problem being solved.
Keywords: metageosystems, spatial data, test sites, territorial descriptors, ensembles
The article presents a methodology for multivariate analysis of intercomponent relationships in geosystems, implemented on the basis of two strategies: the use of numerical, including statistical, methods to assess the degree of influence of various parameters on the state of the geosystem model and the use of simulation to assess the dynamic properties of geosystems and solve the problem of spatial forecasting . The algorithm for assessing the importance of geosystemic parameters makes it possible to evaluate the impact of various parameters on the target indicator in the framework of solving design problems. The concept of building simulation systems for solving the problem of assessing the dynamics and forecasting the development of metageosystems is characterized.
Keywords: metageosystems, spatial data, modeling, intercomponent connections