MACHINE LEARNING APPROACH FOR CORN NITROGEN RECOMMENDATION

• DOI: 10.22533/at.ed.3052302083

• Palavras-chave: Sensor ativo, Random Forest, sensoriamento remoto, milho, estimativa de produtividade

• Keywords: Active sensor, Random Forest, remote sensing, corn, yield estimate

• Abstract:

  Nitrogen (N) fertilizer recommendation tools are vital to precise agricultural management. The objectives of this research were to determine how many variables and remote sensor data are needed to prescribe N fertilizer in corn (Zea mays L.), PFP (partial factor productivity), and yield integrating remote sensing and soil sensor technologies. The variables of this work were NIR, Red, Red-Edge wavelengths, plant height, canopy temperature, LAI (leaf area index), and apparent soil electrical conductivity (ECa). Random Forest Classifier was used to select the best input to estimate N rates, PFP, and corn yield. A confusion matrix was used to identify the accuracy of the Random Forest Classifier to detect the best inputs to estimate for which input we evaluated in this work. According to Random Forest, the best inputs to estimate the N rate and PFP were Red-Edge, Red, and NIR wavelengths, plant height, and canopy temperature. For estimate corn yield were: NIR wavelengths, N rates, plant height, Red-Edge, and canopy temperature.