Prediction of Soybean Yield using Self-normalizing Neural Networks

Abstract

Nowadays, agriculture around the world is facing severe challenges because of global warming and rapid population growth. In order to maximize the agricultural production and minimize the environmental degradation at the same time, careful land-use planning and crop selection come to be crucial, where accurate crop yield prediction plays a key role. This research applies an emerging Deep Learning architecture called Self-normalizing neural networks (SNN) for yield prediction of the soybean, using numerical data obtained from official statistics to ensure high reliability and availability of data. Plentiful work has been done to improve the performance, including careful parameter tuning and application of early stopping and the learning rate scheduler. This study conducts an experiment to evaluate the performance of the model, and the results show that SNN can achieve a lower prediction error with sufficiently large training data compared with traditional machine learning methods, such as Support Vector Regression, and other Deep Learning techniques, such as Batch Normalization.