The market value of fresh citrus fruit is greatly influenced by the internal and external fruit qualities, such as peel color, total soluble solids (TSS), titratable acid (TA), and fruit size. Abundance or deficiency of mineral nutrients in citrus trees are among the most important key factors that affect fruit qualities. Various regression models using leaf nutrient parameters and quality indices have been suggested, but their accuracy and generalization performance in estimating fruit quality remain insufficient. In this research, we used both artificial neural network models (ANN) and a multiple linear regression model to explore the effects of leaf nutrient concentration on citrus fruit quality. For ANN models, we applied two transfer functions and five different training functions to establish the model with best prediction accuracy using TensorFlow framework through Python software. The models were evaluated using statistical performance evaluation criteria including the coefficient of determination (R2), root mean square error (RMSE), mean absolute error (MAE), and mean absolute relative error (MARE). Sensitivity analysis of the prediction models was conducted to discern the significant contribution of leaf mineral nutrients to the respective fruit quality parameters. The response surface analysis determined the optimal range of these mineral elements, which is critical for guiding precision fertilization in fresh market citrus fruit for improving fruit quality. Comprehensive results will be presented during the conference. Keywords: artificial neural network, fruit quality, citrus, mineral nutrients, sensitivity analysis
