Maize yield prediction using LAOA-optimized DeepSCNs

Abstract Maize yield is influenced by various environmental and management factors. Accurate yield prediction plays a key role in optimizing agricultural practices, enhancing production efficiency, and achieving sustainable development. To improve maize yield prediction, we identified key factors such as temperature, humidity, soil moisture, and light intensity using environmental and historical yield data from multiple locations. Based on these factors, we developed a maize yield prediction model using the Improved Arithmetic Optimization Algorithm Optimizing Deep Stochastic Configuration Networks (LAOA-DeepSCNs). First, we trained and tested the LAOA-DeepSCNs model using historical planting data. To validate the model’s performance, we compared its prediction results with those of BP, SVR, and DeepSCNs models. The results show that the LAOA-DeepSCNs model outperforms the other three models in prediction accuracy. Finally, we applied the LAOA-DeepSCNs maize yield prediction model to a real agricultural production environment. The results indicated that within a prediction error range of ± 5% and ± 10%, the hit rates were 90.6% and 92.3%, respectively, meeting the practical requirements of maize production. The results demonstrate that the LAOA-DeepSCNs prediction model significantly enhances the accuracy of maize yield predictions. This model provides a scientific basis for decision-making by farmers and agricultural managers, optimizes field management, improves planting efficiency, reduces production costs, and supports the realization of precision agriculture.