Machine learning for chemical-humus correlation in soil

Abstract

This article investigates the dependency of the quantitative content of humus in soil on phosphate (P2O5), potassium oxide (K2O), hydrolytic acid, as well as the pH value in aqueous and saline environments through machine learning. Linear regression was chosen as the primary model. The mean absolute error (MAE) was found to be 0.517, mean squared error (MSE) – 0.460, and the coefficient of determination after cross-validation reached 0.685. The search for the most significant covariate among the listed ones identified hydrolytic acid as the most impactful due to its influence on microbial activity in the soil and metabolism.