Prediction of sunflower grain yield under normal and salinity stress by RBF, MLP and, CNN models

Sunflower is one of the most valuable oilseeds in the world due to its high-quality oil and wide adaptation to climatic and soil conditions. Salinity is one of the most harmful environmental stresses and severely reduces the yield of crops. In the present study, the effectiveness of multiple regression techniques, convolutional neural network (CNN), and artificial neural network (ANN) are investigated using regression results as input variables in the estimation of sunflower grain yield under normal and salinity conditions, separately. Then the most important parameters identified in two conditions (head diameter, plant height, and weight of five seeds) were used in the CNN model to predict grain yield for the time when we do not know the growth conditions of the plant. The fitted model had R-2 = 0.914, MAPE = 4.95, MAE = 0.163, and RMSE = 1.699. The results showed that the CNN model provides the best estimation for sunflower grain yield compared to the ANN and multiple regression models. Sensitivity analysis showed that head diameter was the most effective trait on sunflower seed yield. Yield estimation with head diameter, identified as the most influential parameter, with the CNN model produced acceptable performance and accuracy.

Automated summary

This study investigates the effectiveness of multiple regression techniques, convolutional neural network (CNN), and artificial neural network (ANN) in estimating the grain yield of sunflower under different conditions. The results show that the CNN model provides the best estimation for sunflower grain yield, with head diameter identified as the most influential factor.