Corn Yield Prediction With Ensemble CNN-DNN

Abstract

We investigate the predictive performance of two novel CNN-DNN machine learning ensemble models in predicting county-level corn yields across the US Corn Belt (12 states). The developed data set is a combination of management, environment, and historical corn yields from 1980 to 2019. Two scenarios for ensemble creation are considered: homogenous and heterogenous ensembles. In homogenous ensembles, the base CNN-DNN models are all the same, but they are generated with a bagging procedure to ensure they exhibit a certain level of diversity. Heterogenous ensembles are created from different base CNN-DNN models which share the same architecture but have different hyperparameters. Three types of ensemble creation methods were used to create several ensembles for either of the scenarios: Basic Ensemble Method (BEM), Generalized Ensemble Method (GEM), and stacked generalized ensembles. Results indicated that both designed ensemble types (heterogenous and homogenous) outperform the ensembles created from five individual ML models (linear regression, LASSO, random forest, XGBoost, and LightGBM). Furthermore, by introducing improvements over the heterogenous ensembles, the homogenous ensembles provide the most accurate yield predictions across US Corn Belt states. This model could make 2019 yield predictions with a root mean square error of 866 kg/ha, equivalent to 8.5% relative root mean square and could successfully explain about 77% of the spatio-temporal variation in the corn grain yields. The significant predictive power of this model can be leveraged for designing a reliable tool for corn yield prediction which will in turn assist agronomic decision makers.

County-scale crop yield prediction by integrating crop simulation with machine learning models

“…Our proposed new model for the same year (2018) had a similar RMSE (1.1 Mg/ha) while providing better interpretability. Shahhosseini et al (2021b) predicted corn yield for the same 12-states in 2019 by implementing an ensemble CNN-DNN model with the weather, soil, and management data. For 2019, the obtained RRMSE was around 8.5%, which is similar to our work (average ensemble model RRMSE of 8.8% and optimized weighted ensemble model RRMSE of 8.9%).…”

“…Residual neural network, a combination of convolutional neural network and recursive neural network (CNN-RNN), was designed to predict corn and soybean yields across the US Corn Belt (Khaki et al, 2020). Shahhosseini et al (2021b) designed an ensemble convolutional neural network-deep neural network (CNN-DNN) architecture to predict corn yield for 12 US Corn Belt states, which had among the lowest prediction error ever reported in the literature (RRMSE of 8.5%). A few recent studies combined probabilistic analysis with neural networks.…”

“…Recently, crop yield prediction models based on neural networks have performed well in terms of prediction accuracy ( Wang et al., 2018 ; Khaki and Wang, 2019 ; Nevavuori et al., 2019 ; Elavarasan and Durairaj Vincent, 2020 ; Shahhosseini et al., 2021b ; Oikonomidis et al., 2022 ). Residual neural network, a combination of convolutional neural network and recursive neural network (CNN-RNN), was designed to predict corn and soybean yields across the US Corn Belt ( Khaki et al., 2020 ).…”

FunSwin: A deep learning method to analysis diabetic retinopathy grade and macular edema risk based on fundus images

“…One is that it is more suitable for convolutional structures by improving the compatibility of function maps and categories, another is that there are no parameters to adjust in the global media collection, meaning that overestimation at the global level can be reduced. It contributes to the achievement of good results in many network structures for medical data ( Bien et al, 2018 ; Valan et al, 2019 ; Shahhosseini et al, 2021 ).…”

Mapping potential malaria vector larval habitats for larval source management: Introduction to multi-model ensembling approaches

“…The reasons for employing ensemble methods in building a model are to enhance the overall performance of the model, minimize the error rate that can be caused by using individual models, and reduce the overall uncertainty of predictions [22,29,30]. There are different ways to ensemble the models, including most-votes, simple average, weighted average (linear or nonlinear), boosting, and stacking [22,24,27,31–33]. In mosquito studies, ensemble modeling has been used to predict the global expansion of Aedes mosquitoes and the invasion of Anopheles stephensi in Africa [26,34,35].…”