Data separation for training the artificial neural network to simulate the spatial distribution of chromium in the surface layer of the soil

Abstract

An algorithm for dividing data into training and test subsamples to simulate the spatial distribution of chromium in the surface layer of the soil using artificial neural networks (ANN) was proposed. The algorithm takes into account the spatial inhomogeneity of the variable being modelled. The data was obtained during the soil screening on the urbanized area in Novyy Urengoy city. A model, which used controlled separation, had shown more accurate results.