Yield prediction and water‐nitrogen management of Chinese jujube based on machine learning

Abstract Jujubes are a crucial characteristic industry in China. Predicting jujube production in various regions of China is significant to developing the jujube industry. This study aims to predict jujube yields across China by machine learning and optimize water‐nitrogen applications to achieve the highest yields. We utilized four machine learning methods (i.e., linear regression, support vector machine, ensemble learning and Gaussian process regression) to create predictive models based on the jujube production, irrigation, fertilization and planting density data sets. The results showed that the Gaussian process regression model best predicted jujube yield by comparing the predicted and measured data. The ensemble learning and Gaussian process regression model best optimized the optimal water and nitrogen application range. On the whole, the Gaussian process regression model is more suitable for yield prediction and water‐nitrogen management. The water‐nitrogen coupling function based on the Gaussian process regression model for predicting jujube yield in Xinjiang, Gansu and Shaanxi was developed to make suitable irrigation and fertilizer regimes. This study can provide a theoretical basis for predicting jujube production and water‐nitrogen management in China.