Prediction of Rice Yield in East China Based on Climate and Agronomic Traits Data Using Artificial Neural Networks and Partial Least Squares Regression

Rice yield is not only influenced by factors of varieties and managements, but also environmental factors. In this study, agronomic trait data rice climate in eastern China were collected, yields predicted using a variety algorithms, including the non-linear tool feed-forward backpropagation neural networks (FFBN) linear model partial least squares regression (PLSR). The results showed that both traits significantly related with yield. PLSR covariates occurred among parameters, modifications should be considered for data-based modelling. FFBN demonstrated better prediction performance than PLSR, which relation coefficient (R2) root mean square error (RMSE) 0.611 vs. 0.374 0.578 0.865 ton/ha data, respectively; 0.742 0.689 0.556 0.608 respectively. When fused R2 RMSE improved to 0.843 0.746 0.440 0.549, optimum architecture consisted one hidden layer 29 neurons. Therefore, algorithm an effective option complex systems production.