APPLICATION OF MACHINE LEARNING MODELS IN MISCANTHUS X GIGANTEUS YIELD ESTIMATION

Miscanthus x giganteus (MxG) is a perennial crop that has high potential for energy production due to its favorable physical and chemical properties and high yield per unit area. To accelerate the process of yield estimation of MxG based on different input parameters, there is a possibility to apply different forms of nonlinear models such as polynomials, artificial neural networks (ANN), support vector machine (SVM) and random forest regression (RFR). In this paper, the aforementioned models were developed in order to predict the yield of Mxg per unit area with respect to the input parameters of plant height and number of shoots. The statistical analysis "goodness of fit" performed, showed high performance in the evaluation; the coefficient of determination (R-2) was used as the main parameter for the effectiveness of the models. Nonlinear models in the form of polynomials (R-2=0.69), SVM (R-2=0.65), RFR (R-2=0.60) and ANN (R-2=0.66) can estimate biomass yield MxG with satisfactory accuracy.