Research on Prediction Intervals of Greenhouse Tomato Yield Based on Improved Particle Swarm Optimization Algorithm

Tomato is the main cultivated crop in China, and greenhouse tomato can improve the quality and yield of tomato by improving the planting environment. Predicting greenhouse tomato yield is an important basis for making production plan, and the accuracy of yield prediction directly affects economic benefits. In order to improve the accuracy of the prediction intervals model of greenhouse tomato yield, a new objective function was proposed to improve the Lower Upper Bound Estimation (LUBE) prediction intervals model. Then, based on the improved LUBE prediction intervals model, the particle swarm optimization algorithm (PSO) was further improved by adaptive inertia weight and escape strategy. Finally, a LUBE prediction intervals model based on the improved PSO algorithm was proposed in this paper. The prediction intervals of greenhouse tomato yield was simulated by using LUBE prediction intervals model, the improved LUBE prediction intervals model and the LUBE prediction intervals model based on the improved PSO algorithm. The experimental results showed that the prediction intervals coverage probability (PICP) of the LUBE prediction intervals model, the improved LUBE prediction intervals model and the LUBE prediction intervals model based on the improved PSO algorithm were 75%, 80% and 80%, respectively, and the prediction intervals normalized average width (PINAW) were 0.3938, 0.3725 and 0.3629, respectively. It can be concluded that the LUBE prediction intervals model based on the improved PSO algorithm has higher prediction accuracy and better fitting ability.