ARTIFICIAL NEURAL NETWORKS FOR PREDICTING SUGARCANE STALK AND TOTAL BIOMASS YIELD BASED ON MICRONUTRIENT RATES APPLIED IN THE PLANTING FURROW AND TO THE LEAVES

Abstract

Sugarcane is pivotal in the global bioeconomy, providing a renewable resource for products such as ethanol, sugar, bioenergy, animal feed, and bioplastics. Its versatility makes it an essential crop for industries seeking sustainable alternatives to fossil fuels. This study presents an advanced approach that uses artificial neural networks (ANNs), specifically a multilayer perceptron model, to accurately estimate sugarcane productivity and biomass. The model incorporates the effects of micronutrient applications, both in the planting furrow and on the leaves, effectively capturing the complex interactions that influence crop yield. During training, the ANN demonstrated high precision, achieving a mean squared error (MSE) of 0.000097 and an R2 of 0.98, closely aligning the predicted outputs with experimental results. In the validation phase, using previously unseen data, it maintained strong performance, with an MSE of 0.0008796. This performance supports the model's ability to generalize beyond the training set, reliably estimating sugarcane yield and biomass under varying conditions. These findings highlight the potential of ANN-based approaches to enhance agricultural management, offering a robust tool to optimize crop performance and improve resource allocation in real-world farming scenarios.